xref: /openbmc/linux/fs/f2fs/file.c (revision 2985bed6)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/file.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <linux/fs.h>
9 #include <linux/f2fs_fs.h>
10 #include <linux/stat.h>
11 #include <linux/buffer_head.h>
12 #include <linux/writeback.h>
13 #include <linux/blkdev.h>
14 #include <linux/falloc.h>
15 #include <linux/types.h>
16 #include <linux/compat.h>
17 #include <linux/uaccess.h>
18 #include <linux/mount.h>
19 #include <linux/pagevec.h>
20 #include <linux/uio.h>
21 #include <linux/uuid.h>
22 #include <linux/file.h>
23 #include <linux/nls.h>
24 
25 #include "f2fs.h"
26 #include "node.h"
27 #include "segment.h"
28 #include "xattr.h"
29 #include "acl.h"
30 #include "gc.h"
31 #include "trace.h"
32 #include <trace/events/f2fs.h>
33 
34 static vm_fault_t f2fs_filemap_fault(struct vm_fault *vmf)
35 {
36 	struct inode *inode = file_inode(vmf->vma->vm_file);
37 	vm_fault_t ret;
38 
39 	down_read(&F2FS_I(inode)->i_mmap_sem);
40 	ret = filemap_fault(vmf);
41 	up_read(&F2FS_I(inode)->i_mmap_sem);
42 
43 	trace_f2fs_filemap_fault(inode, vmf->pgoff, (unsigned long)ret);
44 
45 	return ret;
46 }
47 
48 static vm_fault_t f2fs_vm_page_mkwrite(struct vm_fault *vmf)
49 {
50 	struct page *page = vmf->page;
51 	struct inode *inode = file_inode(vmf->vma->vm_file);
52 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
53 	struct dnode_of_data dn;
54 	bool need_alloc = true;
55 	int err = 0;
56 
57 	if (unlikely(f2fs_cp_error(sbi))) {
58 		err = -EIO;
59 		goto err;
60 	}
61 
62 	if (!f2fs_is_checkpoint_ready(sbi)) {
63 		err = -ENOSPC;
64 		goto err;
65 	}
66 
67 #ifdef CONFIG_F2FS_FS_COMPRESSION
68 	if (f2fs_compressed_file(inode)) {
69 		int ret = f2fs_is_compressed_cluster(inode, page->index);
70 
71 		if (ret < 0) {
72 			err = ret;
73 			goto err;
74 		} else if (ret) {
75 			if (ret < F2FS_I(inode)->i_cluster_size) {
76 				err = -EAGAIN;
77 				goto err;
78 			}
79 			need_alloc = false;
80 		}
81 	}
82 #endif
83 	/* should do out of any locked page */
84 	if (need_alloc)
85 		f2fs_balance_fs(sbi, true);
86 
87 	sb_start_pagefault(inode->i_sb);
88 
89 	f2fs_bug_on(sbi, f2fs_has_inline_data(inode));
90 
91 	file_update_time(vmf->vma->vm_file);
92 	down_read(&F2FS_I(inode)->i_mmap_sem);
93 	lock_page(page);
94 	if (unlikely(page->mapping != inode->i_mapping ||
95 			page_offset(page) > i_size_read(inode) ||
96 			!PageUptodate(page))) {
97 		unlock_page(page);
98 		err = -EFAULT;
99 		goto out_sem;
100 	}
101 
102 	if (need_alloc) {
103 		/* block allocation */
104 		__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true);
105 		set_new_dnode(&dn, inode, NULL, NULL, 0);
106 		err = f2fs_get_block(&dn, page->index);
107 		f2fs_put_dnode(&dn);
108 		__do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false);
109 		if (err) {
110 			unlock_page(page);
111 			goto out_sem;
112 		}
113 	}
114 
115 	/* fill the page */
116 	f2fs_wait_on_page_writeback(page, DATA, false, true);
117 
118 	/* wait for GCed page writeback via META_MAPPING */
119 	f2fs_wait_on_block_writeback(inode, dn.data_blkaddr);
120 
121 	/*
122 	 * check to see if the page is mapped already (no holes)
123 	 */
124 	if (PageMappedToDisk(page))
125 		goto out_sem;
126 
127 	/* page is wholly or partially inside EOF */
128 	if (((loff_t)(page->index + 1) << PAGE_SHIFT) >
129 						i_size_read(inode)) {
130 		loff_t offset;
131 
132 		offset = i_size_read(inode) & ~PAGE_MASK;
133 		zero_user_segment(page, offset, PAGE_SIZE);
134 	}
135 	set_page_dirty(page);
136 	if (!PageUptodate(page))
137 		SetPageUptodate(page);
138 
139 	f2fs_update_iostat(sbi, APP_MAPPED_IO, F2FS_BLKSIZE);
140 	f2fs_update_time(sbi, REQ_TIME);
141 
142 	trace_f2fs_vm_page_mkwrite(page, DATA);
143 out_sem:
144 	up_read(&F2FS_I(inode)->i_mmap_sem);
145 
146 	sb_end_pagefault(inode->i_sb);
147 err:
148 	return block_page_mkwrite_return(err);
149 }
150 
151 static const struct vm_operations_struct f2fs_file_vm_ops = {
152 	.fault		= f2fs_filemap_fault,
153 	.map_pages	= filemap_map_pages,
154 	.page_mkwrite	= f2fs_vm_page_mkwrite,
155 };
156 
157 static int get_parent_ino(struct inode *inode, nid_t *pino)
158 {
159 	struct dentry *dentry;
160 
161 	inode = igrab(inode);
162 	dentry = d_find_any_alias(inode);
163 	iput(inode);
164 	if (!dentry)
165 		return 0;
166 
167 	*pino = parent_ino(dentry);
168 	dput(dentry);
169 	return 1;
170 }
171 
172 static inline enum cp_reason_type need_do_checkpoint(struct inode *inode)
173 {
174 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
175 	enum cp_reason_type cp_reason = CP_NO_NEEDED;
176 
177 	if (!S_ISREG(inode->i_mode))
178 		cp_reason = CP_NON_REGULAR;
179 	else if (f2fs_compressed_file(inode))
180 		cp_reason = CP_COMPRESSED;
181 	else if (inode->i_nlink != 1)
182 		cp_reason = CP_HARDLINK;
183 	else if (is_sbi_flag_set(sbi, SBI_NEED_CP))
184 		cp_reason = CP_SB_NEED_CP;
185 	else if (file_wrong_pino(inode))
186 		cp_reason = CP_WRONG_PINO;
187 	else if (!f2fs_space_for_roll_forward(sbi))
188 		cp_reason = CP_NO_SPC_ROLL;
189 	else if (!f2fs_is_checkpointed_node(sbi, F2FS_I(inode)->i_pino))
190 		cp_reason = CP_NODE_NEED_CP;
191 	else if (test_opt(sbi, FASTBOOT))
192 		cp_reason = CP_FASTBOOT_MODE;
193 	else if (F2FS_OPTION(sbi).active_logs == 2)
194 		cp_reason = CP_SPEC_LOG_NUM;
195 	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT &&
196 		f2fs_need_dentry_mark(sbi, inode->i_ino) &&
197 		f2fs_exist_written_data(sbi, F2FS_I(inode)->i_pino,
198 							TRANS_DIR_INO))
199 		cp_reason = CP_RECOVER_DIR;
200 
201 	return cp_reason;
202 }
203 
204 static bool need_inode_page_update(struct f2fs_sb_info *sbi, nid_t ino)
205 {
206 	struct page *i = find_get_page(NODE_MAPPING(sbi), ino);
207 	bool ret = false;
208 	/* But we need to avoid that there are some inode updates */
209 	if ((i && PageDirty(i)) || f2fs_need_inode_block_update(sbi, ino))
210 		ret = true;
211 	f2fs_put_page(i, 0);
212 	return ret;
213 }
214 
215 static void try_to_fix_pino(struct inode *inode)
216 {
217 	struct f2fs_inode_info *fi = F2FS_I(inode);
218 	nid_t pino;
219 
220 	down_write(&fi->i_sem);
221 	if (file_wrong_pino(inode) && inode->i_nlink == 1 &&
222 			get_parent_ino(inode, &pino)) {
223 		f2fs_i_pino_write(inode, pino);
224 		file_got_pino(inode);
225 	}
226 	up_write(&fi->i_sem);
227 }
228 
229 static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
230 						int datasync, bool atomic)
231 {
232 	struct inode *inode = file->f_mapping->host;
233 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
234 	nid_t ino = inode->i_ino;
235 	int ret = 0;
236 	enum cp_reason_type cp_reason = 0;
237 	struct writeback_control wbc = {
238 		.sync_mode = WB_SYNC_ALL,
239 		.nr_to_write = LONG_MAX,
240 		.for_reclaim = 0,
241 	};
242 	unsigned int seq_id = 0;
243 
244 	if (unlikely(f2fs_readonly(inode->i_sb) ||
245 				is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
246 		return 0;
247 
248 	trace_f2fs_sync_file_enter(inode);
249 
250 	if (S_ISDIR(inode->i_mode))
251 		goto go_write;
252 
253 	/* if fdatasync is triggered, let's do in-place-update */
254 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
255 		set_inode_flag(inode, FI_NEED_IPU);
256 	ret = file_write_and_wait_range(file, start, end);
257 	clear_inode_flag(inode, FI_NEED_IPU);
258 
259 	if (ret) {
260 		trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
261 		return ret;
262 	}
263 
264 	/* if the inode is dirty, let's recover all the time */
265 	if (!f2fs_skip_inode_update(inode, datasync)) {
266 		f2fs_write_inode(inode, NULL);
267 		goto go_write;
268 	}
269 
270 	/*
271 	 * if there is no written data, don't waste time to write recovery info.
272 	 */
273 	if (!is_inode_flag_set(inode, FI_APPEND_WRITE) &&
274 			!f2fs_exist_written_data(sbi, ino, APPEND_INO)) {
275 
276 		/* it may call write_inode just prior to fsync */
277 		if (need_inode_page_update(sbi, ino))
278 			goto go_write;
279 
280 		if (is_inode_flag_set(inode, FI_UPDATE_WRITE) ||
281 				f2fs_exist_written_data(sbi, ino, UPDATE_INO))
282 			goto flush_out;
283 		goto out;
284 	}
285 go_write:
286 	/*
287 	 * Both of fdatasync() and fsync() are able to be recovered from
288 	 * sudden-power-off.
289 	 */
290 	down_read(&F2FS_I(inode)->i_sem);
291 	cp_reason = need_do_checkpoint(inode);
292 	up_read(&F2FS_I(inode)->i_sem);
293 
294 	if (cp_reason) {
295 		/* all the dirty node pages should be flushed for POR */
296 		ret = f2fs_sync_fs(inode->i_sb, 1);
297 
298 		/*
299 		 * We've secured consistency through sync_fs. Following pino
300 		 * will be used only for fsynced inodes after checkpoint.
301 		 */
302 		try_to_fix_pino(inode);
303 		clear_inode_flag(inode, FI_APPEND_WRITE);
304 		clear_inode_flag(inode, FI_UPDATE_WRITE);
305 		goto out;
306 	}
307 sync_nodes:
308 	atomic_inc(&sbi->wb_sync_req[NODE]);
309 	ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id);
310 	atomic_dec(&sbi->wb_sync_req[NODE]);
311 	if (ret)
312 		goto out;
313 
314 	/* if cp_error was enabled, we should avoid infinite loop */
315 	if (unlikely(f2fs_cp_error(sbi))) {
316 		ret = -EIO;
317 		goto out;
318 	}
319 
320 	if (f2fs_need_inode_block_update(sbi, ino)) {
321 		f2fs_mark_inode_dirty_sync(inode, true);
322 		f2fs_write_inode(inode, NULL);
323 		goto sync_nodes;
324 	}
325 
326 	/*
327 	 * If it's atomic_write, it's just fine to keep write ordering. So
328 	 * here we don't need to wait for node write completion, since we use
329 	 * node chain which serializes node blocks. If one of node writes are
330 	 * reordered, we can see simply broken chain, resulting in stopping
331 	 * roll-forward recovery. It means we'll recover all or none node blocks
332 	 * given fsync mark.
333 	 */
334 	if (!atomic) {
335 		ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id);
336 		if (ret)
337 			goto out;
338 	}
339 
340 	/* once recovery info is written, don't need to tack this */
341 	f2fs_remove_ino_entry(sbi, ino, APPEND_INO);
342 	clear_inode_flag(inode, FI_APPEND_WRITE);
343 flush_out:
344 	if (!atomic && F2FS_OPTION(sbi).fsync_mode != FSYNC_MODE_NOBARRIER)
345 		ret = f2fs_issue_flush(sbi, inode->i_ino);
346 	if (!ret) {
347 		f2fs_remove_ino_entry(sbi, ino, UPDATE_INO);
348 		clear_inode_flag(inode, FI_UPDATE_WRITE);
349 		f2fs_remove_ino_entry(sbi, ino, FLUSH_INO);
350 	}
351 	f2fs_update_time(sbi, REQ_TIME);
352 out:
353 	trace_f2fs_sync_file_exit(inode, cp_reason, datasync, ret);
354 	f2fs_trace_ios(NULL, 1);
355 	return ret;
356 }
357 
358 int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
359 {
360 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(file)))))
361 		return -EIO;
362 	return f2fs_do_sync_file(file, start, end, datasync, false);
363 }
364 
365 static pgoff_t __get_first_dirty_index(struct address_space *mapping,
366 						pgoff_t pgofs, int whence)
367 {
368 	struct page *page;
369 	int nr_pages;
370 
371 	if (whence != SEEK_DATA)
372 		return 0;
373 
374 	/* find first dirty page index */
375 	nr_pages = find_get_pages_tag(mapping, &pgofs, PAGECACHE_TAG_DIRTY,
376 				      1, &page);
377 	if (!nr_pages)
378 		return ULONG_MAX;
379 	pgofs = page->index;
380 	put_page(page);
381 	return pgofs;
382 }
383 
384 static bool __found_offset(struct f2fs_sb_info *sbi, block_t blkaddr,
385 				pgoff_t dirty, pgoff_t pgofs, int whence)
386 {
387 	switch (whence) {
388 	case SEEK_DATA:
389 		if ((blkaddr == NEW_ADDR && dirty == pgofs) ||
390 			__is_valid_data_blkaddr(blkaddr))
391 			return true;
392 		break;
393 	case SEEK_HOLE:
394 		if (blkaddr == NULL_ADDR)
395 			return true;
396 		break;
397 	}
398 	return false;
399 }
400 
401 static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
402 {
403 	struct inode *inode = file->f_mapping->host;
404 	loff_t maxbytes = inode->i_sb->s_maxbytes;
405 	struct dnode_of_data dn;
406 	pgoff_t pgofs, end_offset, dirty;
407 	loff_t data_ofs = offset;
408 	loff_t isize;
409 	int err = 0;
410 
411 	inode_lock(inode);
412 
413 	isize = i_size_read(inode);
414 	if (offset >= isize)
415 		goto fail;
416 
417 	/* handle inline data case */
418 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode)) {
419 		if (whence == SEEK_HOLE)
420 			data_ofs = isize;
421 		goto found;
422 	}
423 
424 	pgofs = (pgoff_t)(offset >> PAGE_SHIFT);
425 
426 	dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
427 
428 	for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
429 		set_new_dnode(&dn, inode, NULL, NULL, 0);
430 		err = f2fs_get_dnode_of_data(&dn, pgofs, LOOKUP_NODE);
431 		if (err && err != -ENOENT) {
432 			goto fail;
433 		} else if (err == -ENOENT) {
434 			/* direct node does not exists */
435 			if (whence == SEEK_DATA) {
436 				pgofs = f2fs_get_next_page_offset(&dn, pgofs);
437 				continue;
438 			} else {
439 				goto found;
440 			}
441 		}
442 
443 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
444 
445 		/* find data/hole in dnode block */
446 		for (; dn.ofs_in_node < end_offset;
447 				dn.ofs_in_node++, pgofs++,
448 				data_ofs = (loff_t)pgofs << PAGE_SHIFT) {
449 			block_t blkaddr;
450 
451 			blkaddr = datablock_addr(dn.inode,
452 					dn.node_page, dn.ofs_in_node);
453 
454 			if (__is_valid_data_blkaddr(blkaddr) &&
455 				!f2fs_is_valid_blkaddr(F2FS_I_SB(inode),
456 					blkaddr, DATA_GENERIC_ENHANCE)) {
457 				f2fs_put_dnode(&dn);
458 				goto fail;
459 			}
460 
461 			if (__found_offset(F2FS_I_SB(inode), blkaddr, dirty,
462 							pgofs, whence)) {
463 				f2fs_put_dnode(&dn);
464 				goto found;
465 			}
466 		}
467 		f2fs_put_dnode(&dn);
468 	}
469 
470 	if (whence == SEEK_DATA)
471 		goto fail;
472 found:
473 	if (whence == SEEK_HOLE && data_ofs > isize)
474 		data_ofs = isize;
475 	inode_unlock(inode);
476 	return vfs_setpos(file, data_ofs, maxbytes);
477 fail:
478 	inode_unlock(inode);
479 	return -ENXIO;
480 }
481 
482 static loff_t f2fs_llseek(struct file *file, loff_t offset, int whence)
483 {
484 	struct inode *inode = file->f_mapping->host;
485 	loff_t maxbytes = inode->i_sb->s_maxbytes;
486 
487 	switch (whence) {
488 	case SEEK_SET:
489 	case SEEK_CUR:
490 	case SEEK_END:
491 		return generic_file_llseek_size(file, offset, whence,
492 						maxbytes, i_size_read(inode));
493 	case SEEK_DATA:
494 	case SEEK_HOLE:
495 		if (offset < 0)
496 			return -ENXIO;
497 		return f2fs_seek_block(file, offset, whence);
498 	}
499 
500 	return -EINVAL;
501 }
502 
503 static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
504 {
505 	struct inode *inode = file_inode(file);
506 	int err;
507 
508 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
509 		return -EIO;
510 
511 	if (!f2fs_is_compress_backend_ready(inode))
512 		return -EOPNOTSUPP;
513 
514 	/* we don't need to use inline_data strictly */
515 	err = f2fs_convert_inline_inode(inode);
516 	if (err)
517 		return err;
518 
519 	file_accessed(file);
520 	vma->vm_ops = &f2fs_file_vm_ops;
521 	set_inode_flag(inode, FI_MMAP_FILE);
522 	return 0;
523 }
524 
525 static int f2fs_file_open(struct inode *inode, struct file *filp)
526 {
527 	int err = fscrypt_file_open(inode, filp);
528 
529 	if (err)
530 		return err;
531 
532 	if (!f2fs_is_compress_backend_ready(inode))
533 		return -EOPNOTSUPP;
534 
535 	err = fsverity_file_open(inode, filp);
536 	if (err)
537 		return err;
538 
539 	filp->f_mode |= FMODE_NOWAIT;
540 
541 	return dquot_file_open(inode, filp);
542 }
543 
544 void f2fs_truncate_data_blocks_range(struct dnode_of_data *dn, int count)
545 {
546 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
547 	struct f2fs_node *raw_node;
548 	int nr_free = 0, ofs = dn->ofs_in_node, len = count;
549 	__le32 *addr;
550 	int base = 0;
551 	bool compressed_cluster = false;
552 	int cluster_index = 0, valid_blocks = 0;
553 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
554 
555 	if (IS_INODE(dn->node_page) && f2fs_has_extra_attr(dn->inode))
556 		base = get_extra_isize(dn->inode);
557 
558 	raw_node = F2FS_NODE(dn->node_page);
559 	addr = blkaddr_in_node(raw_node) + base + ofs;
560 
561 	/* Assumption: truncateion starts with cluster */
562 	for (; count > 0; count--, addr++, dn->ofs_in_node++, cluster_index++) {
563 		block_t blkaddr = le32_to_cpu(*addr);
564 
565 		if (f2fs_compressed_file(dn->inode) &&
566 					!(cluster_index & (cluster_size - 1))) {
567 			if (compressed_cluster)
568 				f2fs_i_compr_blocks_update(dn->inode,
569 							valid_blocks, false);
570 			compressed_cluster = (blkaddr == COMPRESS_ADDR);
571 			valid_blocks = 0;
572 		}
573 
574 		if (blkaddr == NULL_ADDR)
575 			continue;
576 
577 		dn->data_blkaddr = NULL_ADDR;
578 		f2fs_set_data_blkaddr(dn);
579 
580 		if (__is_valid_data_blkaddr(blkaddr)) {
581 			if (!f2fs_is_valid_blkaddr(sbi, blkaddr,
582 					DATA_GENERIC_ENHANCE))
583 				continue;
584 			if (compressed_cluster)
585 				valid_blocks++;
586 		}
587 
588 		if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
589 			clear_inode_flag(dn->inode, FI_FIRST_BLOCK_WRITTEN);
590 
591 		f2fs_invalidate_blocks(sbi, blkaddr);
592 		nr_free++;
593 	}
594 
595 	if (compressed_cluster)
596 		f2fs_i_compr_blocks_update(dn->inode, valid_blocks, false);
597 
598 	if (nr_free) {
599 		pgoff_t fofs;
600 		/*
601 		 * once we invalidate valid blkaddr in range [ofs, ofs + count],
602 		 * we will invalidate all blkaddr in the whole range.
603 		 */
604 		fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page),
605 							dn->inode) + ofs;
606 		f2fs_update_extent_cache_range(dn, fofs, 0, len);
607 		dec_valid_block_count(sbi, dn->inode, nr_free);
608 	}
609 	dn->ofs_in_node = ofs;
610 
611 	f2fs_update_time(sbi, REQ_TIME);
612 	trace_f2fs_truncate_data_blocks_range(dn->inode, dn->nid,
613 					 dn->ofs_in_node, nr_free);
614 }
615 
616 void f2fs_truncate_data_blocks(struct dnode_of_data *dn)
617 {
618 	f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
619 }
620 
621 static int truncate_partial_data_page(struct inode *inode, u64 from,
622 								bool cache_only)
623 {
624 	loff_t offset = from & (PAGE_SIZE - 1);
625 	pgoff_t index = from >> PAGE_SHIFT;
626 	struct address_space *mapping = inode->i_mapping;
627 	struct page *page;
628 
629 	if (!offset && !cache_only)
630 		return 0;
631 
632 	if (cache_only) {
633 		page = find_lock_page(mapping, index);
634 		if (page && PageUptodate(page))
635 			goto truncate_out;
636 		f2fs_put_page(page, 1);
637 		return 0;
638 	}
639 
640 	if (f2fs_compressed_file(inode))
641 		return 0;
642 
643 	page = f2fs_get_lock_data_page(inode, index, true);
644 	if (IS_ERR(page))
645 		return PTR_ERR(page) == -ENOENT ? 0 : PTR_ERR(page);
646 truncate_out:
647 	f2fs_wait_on_page_writeback(page, DATA, true, true);
648 	zero_user(page, offset, PAGE_SIZE - offset);
649 
650 	/* An encrypted inode should have a key and truncate the last page. */
651 	f2fs_bug_on(F2FS_I_SB(inode), cache_only && IS_ENCRYPTED(inode));
652 	if (!cache_only)
653 		set_page_dirty(page);
654 	f2fs_put_page(page, 1);
655 	return 0;
656 }
657 
658 static int do_truncate_blocks(struct inode *inode, u64 from, bool lock)
659 {
660 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
661 	struct dnode_of_data dn;
662 	pgoff_t free_from;
663 	int count = 0, err = 0;
664 	struct page *ipage;
665 	bool truncate_page = false;
666 
667 	trace_f2fs_truncate_blocks_enter(inode, from);
668 
669 	free_from = (pgoff_t)F2FS_BLK_ALIGN(from);
670 
671 	if (free_from >= sbi->max_file_blocks)
672 		goto free_partial;
673 
674 	if (lock)
675 		f2fs_lock_op(sbi);
676 
677 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
678 	if (IS_ERR(ipage)) {
679 		err = PTR_ERR(ipage);
680 		goto out;
681 	}
682 
683 	if (f2fs_has_inline_data(inode)) {
684 		f2fs_truncate_inline_inode(inode, ipage, from);
685 		f2fs_put_page(ipage, 1);
686 		truncate_page = true;
687 		goto out;
688 	}
689 
690 	set_new_dnode(&dn, inode, ipage, NULL, 0);
691 	err = f2fs_get_dnode_of_data(&dn, free_from, LOOKUP_NODE_RA);
692 	if (err) {
693 		if (err == -ENOENT)
694 			goto free_next;
695 		goto out;
696 	}
697 
698 	count = ADDRS_PER_PAGE(dn.node_page, inode);
699 
700 	count -= dn.ofs_in_node;
701 	f2fs_bug_on(sbi, count < 0);
702 
703 	if (dn.ofs_in_node || IS_INODE(dn.node_page)) {
704 		f2fs_truncate_data_blocks_range(&dn, count);
705 		free_from += count;
706 	}
707 
708 	f2fs_put_dnode(&dn);
709 free_next:
710 	err = f2fs_truncate_inode_blocks(inode, free_from);
711 out:
712 	if (lock)
713 		f2fs_unlock_op(sbi);
714 free_partial:
715 	/* lastly zero out the first data page */
716 	if (!err)
717 		err = truncate_partial_data_page(inode, from, truncate_page);
718 
719 	trace_f2fs_truncate_blocks_exit(inode, err);
720 	return err;
721 }
722 
723 int f2fs_truncate_blocks(struct inode *inode, u64 from, bool lock)
724 {
725 	u64 free_from = from;
726 
727 	/*
728 	 * for compressed file, only support cluster size
729 	 * aligned truncation.
730 	 */
731 	if (f2fs_compressed_file(inode)) {
732 		size_t cluster_shift = PAGE_SHIFT +
733 					F2FS_I(inode)->i_log_cluster_size;
734 		size_t cluster_mask = (1 << cluster_shift) - 1;
735 
736 		free_from = from >> cluster_shift;
737 		if (from & cluster_mask)
738 			free_from++;
739 		free_from <<= cluster_shift;
740 	}
741 
742 	return do_truncate_blocks(inode, free_from, lock);
743 }
744 
745 int f2fs_truncate(struct inode *inode)
746 {
747 	int err;
748 
749 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
750 		return -EIO;
751 
752 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
753 				S_ISLNK(inode->i_mode)))
754 		return 0;
755 
756 	trace_f2fs_truncate(inode);
757 
758 	if (time_to_inject(F2FS_I_SB(inode), FAULT_TRUNCATE)) {
759 		f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_TRUNCATE);
760 		return -EIO;
761 	}
762 
763 	/* we should check inline_data size */
764 	if (!f2fs_may_inline_data(inode)) {
765 		err = f2fs_convert_inline_inode(inode);
766 		if (err)
767 			return err;
768 	}
769 
770 	err = f2fs_truncate_blocks(inode, i_size_read(inode), true);
771 	if (err)
772 		return err;
773 
774 	inode->i_mtime = inode->i_ctime = current_time(inode);
775 	f2fs_mark_inode_dirty_sync(inode, false);
776 	return 0;
777 }
778 
779 int f2fs_getattr(const struct path *path, struct kstat *stat,
780 		 u32 request_mask, unsigned int query_flags)
781 {
782 	struct inode *inode = d_inode(path->dentry);
783 	struct f2fs_inode_info *fi = F2FS_I(inode);
784 	struct f2fs_inode *ri;
785 	unsigned int flags;
786 
787 	if (f2fs_has_extra_attr(inode) &&
788 			f2fs_sb_has_inode_crtime(F2FS_I_SB(inode)) &&
789 			F2FS_FITS_IN_INODE(ri, fi->i_extra_isize, i_crtime)) {
790 		stat->result_mask |= STATX_BTIME;
791 		stat->btime.tv_sec = fi->i_crtime.tv_sec;
792 		stat->btime.tv_nsec = fi->i_crtime.tv_nsec;
793 	}
794 
795 	flags = fi->i_flags;
796 	if (flags & F2FS_APPEND_FL)
797 		stat->attributes |= STATX_ATTR_APPEND;
798 	if (IS_ENCRYPTED(inode))
799 		stat->attributes |= STATX_ATTR_ENCRYPTED;
800 	if (flags & F2FS_IMMUTABLE_FL)
801 		stat->attributes |= STATX_ATTR_IMMUTABLE;
802 	if (flags & F2FS_NODUMP_FL)
803 		stat->attributes |= STATX_ATTR_NODUMP;
804 	if (IS_VERITY(inode))
805 		stat->attributes |= STATX_ATTR_VERITY;
806 
807 	stat->attributes_mask |= (STATX_ATTR_APPEND |
808 				  STATX_ATTR_ENCRYPTED |
809 				  STATX_ATTR_IMMUTABLE |
810 				  STATX_ATTR_NODUMP |
811 				  STATX_ATTR_VERITY);
812 
813 	generic_fillattr(inode, stat);
814 
815 	/* we need to show initial sectors used for inline_data/dentries */
816 	if ((S_ISREG(inode->i_mode) && f2fs_has_inline_data(inode)) ||
817 					f2fs_has_inline_dentry(inode))
818 		stat->blocks += (stat->size + 511) >> 9;
819 
820 	return 0;
821 }
822 
823 #ifdef CONFIG_F2FS_FS_POSIX_ACL
824 static void __setattr_copy(struct inode *inode, const struct iattr *attr)
825 {
826 	unsigned int ia_valid = attr->ia_valid;
827 
828 	if (ia_valid & ATTR_UID)
829 		inode->i_uid = attr->ia_uid;
830 	if (ia_valid & ATTR_GID)
831 		inode->i_gid = attr->ia_gid;
832 	if (ia_valid & ATTR_ATIME)
833 		inode->i_atime = attr->ia_atime;
834 	if (ia_valid & ATTR_MTIME)
835 		inode->i_mtime = attr->ia_mtime;
836 	if (ia_valid & ATTR_CTIME)
837 		inode->i_ctime = attr->ia_ctime;
838 	if (ia_valid & ATTR_MODE) {
839 		umode_t mode = attr->ia_mode;
840 
841 		if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
842 			mode &= ~S_ISGID;
843 		set_acl_inode(inode, mode);
844 	}
845 }
846 #else
847 #define __setattr_copy setattr_copy
848 #endif
849 
850 int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
851 {
852 	struct inode *inode = d_inode(dentry);
853 	int err;
854 
855 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
856 		return -EIO;
857 
858 	if ((attr->ia_valid & ATTR_SIZE) &&
859 		!f2fs_is_compress_backend_ready(inode))
860 		return -EOPNOTSUPP;
861 
862 	err = setattr_prepare(dentry, attr);
863 	if (err)
864 		return err;
865 
866 	err = fscrypt_prepare_setattr(dentry, attr);
867 	if (err)
868 		return err;
869 
870 	err = fsverity_prepare_setattr(dentry, attr);
871 	if (err)
872 		return err;
873 
874 	if (is_quota_modification(inode, attr)) {
875 		err = dquot_initialize(inode);
876 		if (err)
877 			return err;
878 	}
879 	if ((attr->ia_valid & ATTR_UID &&
880 		!uid_eq(attr->ia_uid, inode->i_uid)) ||
881 		(attr->ia_valid & ATTR_GID &&
882 		!gid_eq(attr->ia_gid, inode->i_gid))) {
883 		f2fs_lock_op(F2FS_I_SB(inode));
884 		err = dquot_transfer(inode, attr);
885 		if (err) {
886 			set_sbi_flag(F2FS_I_SB(inode),
887 					SBI_QUOTA_NEED_REPAIR);
888 			f2fs_unlock_op(F2FS_I_SB(inode));
889 			return err;
890 		}
891 		/*
892 		 * update uid/gid under lock_op(), so that dquot and inode can
893 		 * be updated atomically.
894 		 */
895 		if (attr->ia_valid & ATTR_UID)
896 			inode->i_uid = attr->ia_uid;
897 		if (attr->ia_valid & ATTR_GID)
898 			inode->i_gid = attr->ia_gid;
899 		f2fs_mark_inode_dirty_sync(inode, true);
900 		f2fs_unlock_op(F2FS_I_SB(inode));
901 	}
902 
903 	if (attr->ia_valid & ATTR_SIZE) {
904 		loff_t old_size = i_size_read(inode);
905 
906 		if (attr->ia_size > MAX_INLINE_DATA(inode)) {
907 			/*
908 			 * should convert inline inode before i_size_write to
909 			 * keep smaller than inline_data size with inline flag.
910 			 */
911 			err = f2fs_convert_inline_inode(inode);
912 			if (err)
913 				return err;
914 		}
915 
916 		down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
917 		down_write(&F2FS_I(inode)->i_mmap_sem);
918 
919 		truncate_setsize(inode, attr->ia_size);
920 
921 		if (attr->ia_size <= old_size)
922 			err = f2fs_truncate(inode);
923 		/*
924 		 * do not trim all blocks after i_size if target size is
925 		 * larger than i_size.
926 		 */
927 		up_write(&F2FS_I(inode)->i_mmap_sem);
928 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
929 		if (err)
930 			return err;
931 
932 		down_write(&F2FS_I(inode)->i_sem);
933 		inode->i_mtime = inode->i_ctime = current_time(inode);
934 		F2FS_I(inode)->last_disk_size = i_size_read(inode);
935 		up_write(&F2FS_I(inode)->i_sem);
936 	}
937 
938 	__setattr_copy(inode, attr);
939 
940 	if (attr->ia_valid & ATTR_MODE) {
941 		err = posix_acl_chmod(inode, f2fs_get_inode_mode(inode));
942 		if (err || is_inode_flag_set(inode, FI_ACL_MODE)) {
943 			inode->i_mode = F2FS_I(inode)->i_acl_mode;
944 			clear_inode_flag(inode, FI_ACL_MODE);
945 		}
946 	}
947 
948 	/* file size may changed here */
949 	f2fs_mark_inode_dirty_sync(inode, true);
950 
951 	/* inode change will produce dirty node pages flushed by checkpoint */
952 	f2fs_balance_fs(F2FS_I_SB(inode), true);
953 
954 	return err;
955 }
956 
957 const struct inode_operations f2fs_file_inode_operations = {
958 	.getattr	= f2fs_getattr,
959 	.setattr	= f2fs_setattr,
960 	.get_acl	= f2fs_get_acl,
961 	.set_acl	= f2fs_set_acl,
962 #ifdef CONFIG_F2FS_FS_XATTR
963 	.listxattr	= f2fs_listxattr,
964 #endif
965 	.fiemap		= f2fs_fiemap,
966 };
967 
968 static int fill_zero(struct inode *inode, pgoff_t index,
969 					loff_t start, loff_t len)
970 {
971 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
972 	struct page *page;
973 
974 	if (!len)
975 		return 0;
976 
977 	f2fs_balance_fs(sbi, true);
978 
979 	f2fs_lock_op(sbi);
980 	page = f2fs_get_new_data_page(inode, NULL, index, false);
981 	f2fs_unlock_op(sbi);
982 
983 	if (IS_ERR(page))
984 		return PTR_ERR(page);
985 
986 	f2fs_wait_on_page_writeback(page, DATA, true, true);
987 	zero_user(page, start, len);
988 	set_page_dirty(page);
989 	f2fs_put_page(page, 1);
990 	return 0;
991 }
992 
993 int f2fs_truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
994 {
995 	int err;
996 
997 	while (pg_start < pg_end) {
998 		struct dnode_of_data dn;
999 		pgoff_t end_offset, count;
1000 
1001 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1002 		err = f2fs_get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
1003 		if (err) {
1004 			if (err == -ENOENT) {
1005 				pg_start = f2fs_get_next_page_offset(&dn,
1006 								pg_start);
1007 				continue;
1008 			}
1009 			return err;
1010 		}
1011 
1012 		end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1013 		count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
1014 
1015 		f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
1016 
1017 		f2fs_truncate_data_blocks_range(&dn, count);
1018 		f2fs_put_dnode(&dn);
1019 
1020 		pg_start += count;
1021 	}
1022 	return 0;
1023 }
1024 
1025 static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
1026 {
1027 	pgoff_t pg_start, pg_end;
1028 	loff_t off_start, off_end;
1029 	int ret;
1030 
1031 	ret = f2fs_convert_inline_inode(inode);
1032 	if (ret)
1033 		return ret;
1034 
1035 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1036 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1037 
1038 	off_start = offset & (PAGE_SIZE - 1);
1039 	off_end = (offset + len) & (PAGE_SIZE - 1);
1040 
1041 	if (pg_start == pg_end) {
1042 		ret = fill_zero(inode, pg_start, off_start,
1043 						off_end - off_start);
1044 		if (ret)
1045 			return ret;
1046 	} else {
1047 		if (off_start) {
1048 			ret = fill_zero(inode, pg_start++, off_start,
1049 						PAGE_SIZE - off_start);
1050 			if (ret)
1051 				return ret;
1052 		}
1053 		if (off_end) {
1054 			ret = fill_zero(inode, pg_end, 0, off_end);
1055 			if (ret)
1056 				return ret;
1057 		}
1058 
1059 		if (pg_start < pg_end) {
1060 			struct address_space *mapping = inode->i_mapping;
1061 			loff_t blk_start, blk_end;
1062 			struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1063 
1064 			f2fs_balance_fs(sbi, true);
1065 
1066 			blk_start = (loff_t)pg_start << PAGE_SHIFT;
1067 			blk_end = (loff_t)pg_end << PAGE_SHIFT;
1068 
1069 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1070 			down_write(&F2FS_I(inode)->i_mmap_sem);
1071 
1072 			truncate_inode_pages_range(mapping, blk_start,
1073 					blk_end - 1);
1074 
1075 			f2fs_lock_op(sbi);
1076 			ret = f2fs_truncate_hole(inode, pg_start, pg_end);
1077 			f2fs_unlock_op(sbi);
1078 
1079 			up_write(&F2FS_I(inode)->i_mmap_sem);
1080 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1081 		}
1082 	}
1083 
1084 	return ret;
1085 }
1086 
1087 static int __read_out_blkaddrs(struct inode *inode, block_t *blkaddr,
1088 				int *do_replace, pgoff_t off, pgoff_t len)
1089 {
1090 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1091 	struct dnode_of_data dn;
1092 	int ret, done, i;
1093 
1094 next_dnode:
1095 	set_new_dnode(&dn, inode, NULL, NULL, 0);
1096 	ret = f2fs_get_dnode_of_data(&dn, off, LOOKUP_NODE_RA);
1097 	if (ret && ret != -ENOENT) {
1098 		return ret;
1099 	} else if (ret == -ENOENT) {
1100 		if (dn.max_level == 0)
1101 			return -ENOENT;
1102 		done = min((pgoff_t)ADDRS_PER_BLOCK(inode) -
1103 						dn.ofs_in_node, len);
1104 		blkaddr += done;
1105 		do_replace += done;
1106 		goto next;
1107 	}
1108 
1109 	done = min((pgoff_t)ADDRS_PER_PAGE(dn.node_page, inode) -
1110 							dn.ofs_in_node, len);
1111 	for (i = 0; i < done; i++, blkaddr++, do_replace++, dn.ofs_in_node++) {
1112 		*blkaddr = datablock_addr(dn.inode,
1113 					dn.node_page, dn.ofs_in_node);
1114 
1115 		if (__is_valid_data_blkaddr(*blkaddr) &&
1116 			!f2fs_is_valid_blkaddr(sbi, *blkaddr,
1117 					DATA_GENERIC_ENHANCE)) {
1118 			f2fs_put_dnode(&dn);
1119 			return -EFSCORRUPTED;
1120 		}
1121 
1122 		if (!f2fs_is_checkpointed_data(sbi, *blkaddr)) {
1123 
1124 			if (test_opt(sbi, LFS)) {
1125 				f2fs_put_dnode(&dn);
1126 				return -EOPNOTSUPP;
1127 			}
1128 
1129 			/* do not invalidate this block address */
1130 			f2fs_update_data_blkaddr(&dn, NULL_ADDR);
1131 			*do_replace = 1;
1132 		}
1133 	}
1134 	f2fs_put_dnode(&dn);
1135 next:
1136 	len -= done;
1137 	off += done;
1138 	if (len)
1139 		goto next_dnode;
1140 	return 0;
1141 }
1142 
1143 static int __roll_back_blkaddrs(struct inode *inode, block_t *blkaddr,
1144 				int *do_replace, pgoff_t off, int len)
1145 {
1146 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1147 	struct dnode_of_data dn;
1148 	int ret, i;
1149 
1150 	for (i = 0; i < len; i++, do_replace++, blkaddr++) {
1151 		if (*do_replace == 0)
1152 			continue;
1153 
1154 		set_new_dnode(&dn, inode, NULL, NULL, 0);
1155 		ret = f2fs_get_dnode_of_data(&dn, off + i, LOOKUP_NODE_RA);
1156 		if (ret) {
1157 			dec_valid_block_count(sbi, inode, 1);
1158 			f2fs_invalidate_blocks(sbi, *blkaddr);
1159 		} else {
1160 			f2fs_update_data_blkaddr(&dn, *blkaddr);
1161 		}
1162 		f2fs_put_dnode(&dn);
1163 	}
1164 	return 0;
1165 }
1166 
1167 static int __clone_blkaddrs(struct inode *src_inode, struct inode *dst_inode,
1168 			block_t *blkaddr, int *do_replace,
1169 			pgoff_t src, pgoff_t dst, pgoff_t len, bool full)
1170 {
1171 	struct f2fs_sb_info *sbi = F2FS_I_SB(src_inode);
1172 	pgoff_t i = 0;
1173 	int ret;
1174 
1175 	while (i < len) {
1176 		if (blkaddr[i] == NULL_ADDR && !full) {
1177 			i++;
1178 			continue;
1179 		}
1180 
1181 		if (do_replace[i] || blkaddr[i] == NULL_ADDR) {
1182 			struct dnode_of_data dn;
1183 			struct node_info ni;
1184 			size_t new_size;
1185 			pgoff_t ilen;
1186 
1187 			set_new_dnode(&dn, dst_inode, NULL, NULL, 0);
1188 			ret = f2fs_get_dnode_of_data(&dn, dst + i, ALLOC_NODE);
1189 			if (ret)
1190 				return ret;
1191 
1192 			ret = f2fs_get_node_info(sbi, dn.nid, &ni);
1193 			if (ret) {
1194 				f2fs_put_dnode(&dn);
1195 				return ret;
1196 			}
1197 
1198 			ilen = min((pgoff_t)
1199 				ADDRS_PER_PAGE(dn.node_page, dst_inode) -
1200 						dn.ofs_in_node, len - i);
1201 			do {
1202 				dn.data_blkaddr = datablock_addr(dn.inode,
1203 						dn.node_page, dn.ofs_in_node);
1204 				f2fs_truncate_data_blocks_range(&dn, 1);
1205 
1206 				if (do_replace[i]) {
1207 					f2fs_i_blocks_write(src_inode,
1208 							1, false, false);
1209 					f2fs_i_blocks_write(dst_inode,
1210 							1, true, false);
1211 					f2fs_replace_block(sbi, &dn, dn.data_blkaddr,
1212 					blkaddr[i], ni.version, true, false);
1213 
1214 					do_replace[i] = 0;
1215 				}
1216 				dn.ofs_in_node++;
1217 				i++;
1218 				new_size = (loff_t)(dst + i) << PAGE_SHIFT;
1219 				if (dst_inode->i_size < new_size)
1220 					f2fs_i_size_write(dst_inode, new_size);
1221 			} while (--ilen && (do_replace[i] || blkaddr[i] == NULL_ADDR));
1222 
1223 			f2fs_put_dnode(&dn);
1224 		} else {
1225 			struct page *psrc, *pdst;
1226 
1227 			psrc = f2fs_get_lock_data_page(src_inode,
1228 							src + i, true);
1229 			if (IS_ERR(psrc))
1230 				return PTR_ERR(psrc);
1231 			pdst = f2fs_get_new_data_page(dst_inode, NULL, dst + i,
1232 								true);
1233 			if (IS_ERR(pdst)) {
1234 				f2fs_put_page(psrc, 1);
1235 				return PTR_ERR(pdst);
1236 			}
1237 			f2fs_copy_page(psrc, pdst);
1238 			set_page_dirty(pdst);
1239 			f2fs_put_page(pdst, 1);
1240 			f2fs_put_page(psrc, 1);
1241 
1242 			ret = f2fs_truncate_hole(src_inode,
1243 						src + i, src + i + 1);
1244 			if (ret)
1245 				return ret;
1246 			i++;
1247 		}
1248 	}
1249 	return 0;
1250 }
1251 
1252 static int __exchange_data_block(struct inode *src_inode,
1253 			struct inode *dst_inode, pgoff_t src, pgoff_t dst,
1254 			pgoff_t len, bool full)
1255 {
1256 	block_t *src_blkaddr;
1257 	int *do_replace;
1258 	pgoff_t olen;
1259 	int ret;
1260 
1261 	while (len) {
1262 		olen = min((pgoff_t)4 * ADDRS_PER_BLOCK(src_inode), len);
1263 
1264 		src_blkaddr = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1265 					array_size(olen, sizeof(block_t)),
1266 					GFP_NOFS);
1267 		if (!src_blkaddr)
1268 			return -ENOMEM;
1269 
1270 		do_replace = f2fs_kvzalloc(F2FS_I_SB(src_inode),
1271 					array_size(olen, sizeof(int)),
1272 					GFP_NOFS);
1273 		if (!do_replace) {
1274 			kvfree(src_blkaddr);
1275 			return -ENOMEM;
1276 		}
1277 
1278 		ret = __read_out_blkaddrs(src_inode, src_blkaddr,
1279 					do_replace, src, olen);
1280 		if (ret)
1281 			goto roll_back;
1282 
1283 		ret = __clone_blkaddrs(src_inode, dst_inode, src_blkaddr,
1284 					do_replace, src, dst, olen, full);
1285 		if (ret)
1286 			goto roll_back;
1287 
1288 		src += olen;
1289 		dst += olen;
1290 		len -= olen;
1291 
1292 		kvfree(src_blkaddr);
1293 		kvfree(do_replace);
1294 	}
1295 	return 0;
1296 
1297 roll_back:
1298 	__roll_back_blkaddrs(src_inode, src_blkaddr, do_replace, src, olen);
1299 	kvfree(src_blkaddr);
1300 	kvfree(do_replace);
1301 	return ret;
1302 }
1303 
1304 static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
1305 {
1306 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1307 	pgoff_t nrpages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1308 	pgoff_t start = offset >> PAGE_SHIFT;
1309 	pgoff_t end = (offset + len) >> PAGE_SHIFT;
1310 	int ret;
1311 
1312 	f2fs_balance_fs(sbi, true);
1313 
1314 	/* avoid gc operation during block exchange */
1315 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1316 	down_write(&F2FS_I(inode)->i_mmap_sem);
1317 
1318 	f2fs_lock_op(sbi);
1319 	f2fs_drop_extent_tree(inode);
1320 	truncate_pagecache(inode, offset);
1321 	ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
1322 	f2fs_unlock_op(sbi);
1323 
1324 	up_write(&F2FS_I(inode)->i_mmap_sem);
1325 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1326 	return ret;
1327 }
1328 
1329 static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
1330 {
1331 	loff_t new_size;
1332 	int ret;
1333 
1334 	if (offset + len >= i_size_read(inode))
1335 		return -EINVAL;
1336 
1337 	/* collapse range should be aligned to block size of f2fs. */
1338 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1339 		return -EINVAL;
1340 
1341 	ret = f2fs_convert_inline_inode(inode);
1342 	if (ret)
1343 		return ret;
1344 
1345 	/* write out all dirty pages from offset */
1346 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1347 	if (ret)
1348 		return ret;
1349 
1350 	ret = f2fs_do_collapse(inode, offset, len);
1351 	if (ret)
1352 		return ret;
1353 
1354 	/* write out all moved pages, if possible */
1355 	down_write(&F2FS_I(inode)->i_mmap_sem);
1356 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1357 	truncate_pagecache(inode, offset);
1358 
1359 	new_size = i_size_read(inode) - len;
1360 	truncate_pagecache(inode, new_size);
1361 
1362 	ret = f2fs_truncate_blocks(inode, new_size, true);
1363 	up_write(&F2FS_I(inode)->i_mmap_sem);
1364 	if (!ret)
1365 		f2fs_i_size_write(inode, new_size);
1366 	return ret;
1367 }
1368 
1369 static int f2fs_do_zero_range(struct dnode_of_data *dn, pgoff_t start,
1370 								pgoff_t end)
1371 {
1372 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
1373 	pgoff_t index = start;
1374 	unsigned int ofs_in_node = dn->ofs_in_node;
1375 	blkcnt_t count = 0;
1376 	int ret;
1377 
1378 	for (; index < end; index++, dn->ofs_in_node++) {
1379 		if (datablock_addr(dn->inode, dn->node_page,
1380 					dn->ofs_in_node) == NULL_ADDR)
1381 			count++;
1382 	}
1383 
1384 	dn->ofs_in_node = ofs_in_node;
1385 	ret = f2fs_reserve_new_blocks(dn, count);
1386 	if (ret)
1387 		return ret;
1388 
1389 	dn->ofs_in_node = ofs_in_node;
1390 	for (index = start; index < end; index++, dn->ofs_in_node++) {
1391 		dn->data_blkaddr = datablock_addr(dn->inode,
1392 					dn->node_page, dn->ofs_in_node);
1393 		/*
1394 		 * f2fs_reserve_new_blocks will not guarantee entire block
1395 		 * allocation.
1396 		 */
1397 		if (dn->data_blkaddr == NULL_ADDR) {
1398 			ret = -ENOSPC;
1399 			break;
1400 		}
1401 		if (dn->data_blkaddr != NEW_ADDR) {
1402 			f2fs_invalidate_blocks(sbi, dn->data_blkaddr);
1403 			dn->data_blkaddr = NEW_ADDR;
1404 			f2fs_set_data_blkaddr(dn);
1405 		}
1406 	}
1407 
1408 	f2fs_update_extent_cache_range(dn, start, 0, index - start);
1409 
1410 	return ret;
1411 }
1412 
1413 static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
1414 								int mode)
1415 {
1416 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1417 	struct address_space *mapping = inode->i_mapping;
1418 	pgoff_t index, pg_start, pg_end;
1419 	loff_t new_size = i_size_read(inode);
1420 	loff_t off_start, off_end;
1421 	int ret = 0;
1422 
1423 	ret = inode_newsize_ok(inode, (len + offset));
1424 	if (ret)
1425 		return ret;
1426 
1427 	ret = f2fs_convert_inline_inode(inode);
1428 	if (ret)
1429 		return ret;
1430 
1431 	ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
1432 	if (ret)
1433 		return ret;
1434 
1435 	pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
1436 	pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
1437 
1438 	off_start = offset & (PAGE_SIZE - 1);
1439 	off_end = (offset + len) & (PAGE_SIZE - 1);
1440 
1441 	if (pg_start == pg_end) {
1442 		ret = fill_zero(inode, pg_start, off_start,
1443 						off_end - off_start);
1444 		if (ret)
1445 			return ret;
1446 
1447 		new_size = max_t(loff_t, new_size, offset + len);
1448 	} else {
1449 		if (off_start) {
1450 			ret = fill_zero(inode, pg_start++, off_start,
1451 						PAGE_SIZE - off_start);
1452 			if (ret)
1453 				return ret;
1454 
1455 			new_size = max_t(loff_t, new_size,
1456 					(loff_t)pg_start << PAGE_SHIFT);
1457 		}
1458 
1459 		for (index = pg_start; index < pg_end;) {
1460 			struct dnode_of_data dn;
1461 			unsigned int end_offset;
1462 			pgoff_t end;
1463 
1464 			down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1465 			down_write(&F2FS_I(inode)->i_mmap_sem);
1466 
1467 			truncate_pagecache_range(inode,
1468 				(loff_t)index << PAGE_SHIFT,
1469 				((loff_t)pg_end << PAGE_SHIFT) - 1);
1470 
1471 			f2fs_lock_op(sbi);
1472 
1473 			set_new_dnode(&dn, inode, NULL, NULL, 0);
1474 			ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
1475 			if (ret) {
1476 				f2fs_unlock_op(sbi);
1477 				up_write(&F2FS_I(inode)->i_mmap_sem);
1478 				up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1479 				goto out;
1480 			}
1481 
1482 			end_offset = ADDRS_PER_PAGE(dn.node_page, inode);
1483 			end = min(pg_end, end_offset - dn.ofs_in_node + index);
1484 
1485 			ret = f2fs_do_zero_range(&dn, index, end);
1486 			f2fs_put_dnode(&dn);
1487 
1488 			f2fs_unlock_op(sbi);
1489 			up_write(&F2FS_I(inode)->i_mmap_sem);
1490 			up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1491 
1492 			f2fs_balance_fs(sbi, dn.node_changed);
1493 
1494 			if (ret)
1495 				goto out;
1496 
1497 			index = end;
1498 			new_size = max_t(loff_t, new_size,
1499 					(loff_t)index << PAGE_SHIFT);
1500 		}
1501 
1502 		if (off_end) {
1503 			ret = fill_zero(inode, pg_end, 0, off_end);
1504 			if (ret)
1505 				goto out;
1506 
1507 			new_size = max_t(loff_t, new_size, offset + len);
1508 		}
1509 	}
1510 
1511 out:
1512 	if (new_size > i_size_read(inode)) {
1513 		if (mode & FALLOC_FL_KEEP_SIZE)
1514 			file_set_keep_isize(inode);
1515 		else
1516 			f2fs_i_size_write(inode, new_size);
1517 	}
1518 	return ret;
1519 }
1520 
1521 static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
1522 {
1523 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1524 	pgoff_t nr, pg_start, pg_end, delta, idx;
1525 	loff_t new_size;
1526 	int ret = 0;
1527 
1528 	new_size = i_size_read(inode) + len;
1529 	ret = inode_newsize_ok(inode, new_size);
1530 	if (ret)
1531 		return ret;
1532 
1533 	if (offset >= i_size_read(inode))
1534 		return -EINVAL;
1535 
1536 	/* insert range should be aligned to block size of f2fs. */
1537 	if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
1538 		return -EINVAL;
1539 
1540 	ret = f2fs_convert_inline_inode(inode);
1541 	if (ret)
1542 		return ret;
1543 
1544 	f2fs_balance_fs(sbi, true);
1545 
1546 	down_write(&F2FS_I(inode)->i_mmap_sem);
1547 	ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
1548 	up_write(&F2FS_I(inode)->i_mmap_sem);
1549 	if (ret)
1550 		return ret;
1551 
1552 	/* write out all dirty pages from offset */
1553 	ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1554 	if (ret)
1555 		return ret;
1556 
1557 	pg_start = offset >> PAGE_SHIFT;
1558 	pg_end = (offset + len) >> PAGE_SHIFT;
1559 	delta = pg_end - pg_start;
1560 	idx = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
1561 
1562 	/* avoid gc operation during block exchange */
1563 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1564 	down_write(&F2FS_I(inode)->i_mmap_sem);
1565 	truncate_pagecache(inode, offset);
1566 
1567 	while (!ret && idx > pg_start) {
1568 		nr = idx - pg_start;
1569 		if (nr > delta)
1570 			nr = delta;
1571 		idx -= nr;
1572 
1573 		f2fs_lock_op(sbi);
1574 		f2fs_drop_extent_tree(inode);
1575 
1576 		ret = __exchange_data_block(inode, inode, idx,
1577 					idx + delta, nr, false);
1578 		f2fs_unlock_op(sbi);
1579 	}
1580 	up_write(&F2FS_I(inode)->i_mmap_sem);
1581 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
1582 
1583 	/* write out all moved pages, if possible */
1584 	down_write(&F2FS_I(inode)->i_mmap_sem);
1585 	filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
1586 	truncate_pagecache(inode, offset);
1587 	up_write(&F2FS_I(inode)->i_mmap_sem);
1588 
1589 	if (!ret)
1590 		f2fs_i_size_write(inode, new_size);
1591 	return ret;
1592 }
1593 
1594 static int expand_inode_data(struct inode *inode, loff_t offset,
1595 					loff_t len, int mode)
1596 {
1597 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1598 	struct f2fs_map_blocks map = { .m_next_pgofs = NULL,
1599 			.m_next_extent = NULL, .m_seg_type = NO_CHECK_TYPE,
1600 			.m_may_create = true };
1601 	pgoff_t pg_end;
1602 	loff_t new_size = i_size_read(inode);
1603 	loff_t off_end;
1604 	int err;
1605 
1606 	err = inode_newsize_ok(inode, (len + offset));
1607 	if (err)
1608 		return err;
1609 
1610 	err = f2fs_convert_inline_inode(inode);
1611 	if (err)
1612 		return err;
1613 
1614 	f2fs_balance_fs(sbi, true);
1615 
1616 	pg_end = ((unsigned long long)offset + len) >> PAGE_SHIFT;
1617 	off_end = (offset + len) & (PAGE_SIZE - 1);
1618 
1619 	map.m_lblk = ((unsigned long long)offset) >> PAGE_SHIFT;
1620 	map.m_len = pg_end - map.m_lblk;
1621 	if (off_end)
1622 		map.m_len++;
1623 
1624 	if (!map.m_len)
1625 		return 0;
1626 
1627 	if (f2fs_is_pinned_file(inode)) {
1628 		block_t len = (map.m_len >> sbi->log_blocks_per_seg) <<
1629 					sbi->log_blocks_per_seg;
1630 		block_t done = 0;
1631 
1632 		if (map.m_len % sbi->blocks_per_seg)
1633 			len += sbi->blocks_per_seg;
1634 
1635 		map.m_len = sbi->blocks_per_seg;
1636 next_alloc:
1637 		if (has_not_enough_free_secs(sbi, 0,
1638 			GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)))) {
1639 			down_write(&sbi->gc_lock);
1640 			err = f2fs_gc(sbi, true, false, NULL_SEGNO);
1641 			if (err && err != -ENODATA && err != -EAGAIN)
1642 				goto out_err;
1643 		}
1644 
1645 		down_write(&sbi->pin_sem);
1646 		map.m_seg_type = CURSEG_COLD_DATA_PINNED;
1647 		f2fs_allocate_new_segments(sbi, CURSEG_COLD_DATA);
1648 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_DIO);
1649 		up_write(&sbi->pin_sem);
1650 
1651 		done += map.m_len;
1652 		len -= map.m_len;
1653 		map.m_lblk += map.m_len;
1654 		if (!err && len)
1655 			goto next_alloc;
1656 
1657 		map.m_len = done;
1658 	} else {
1659 		err = f2fs_map_blocks(inode, &map, 1, F2FS_GET_BLOCK_PRE_AIO);
1660 	}
1661 out_err:
1662 	if (err) {
1663 		pgoff_t last_off;
1664 
1665 		if (!map.m_len)
1666 			return err;
1667 
1668 		last_off = map.m_lblk + map.m_len - 1;
1669 
1670 		/* update new size to the failed position */
1671 		new_size = (last_off == pg_end) ? offset + len :
1672 					(loff_t)(last_off + 1) << PAGE_SHIFT;
1673 	} else {
1674 		new_size = ((loff_t)pg_end << PAGE_SHIFT) + off_end;
1675 	}
1676 
1677 	if (new_size > i_size_read(inode)) {
1678 		if (mode & FALLOC_FL_KEEP_SIZE)
1679 			file_set_keep_isize(inode);
1680 		else
1681 			f2fs_i_size_write(inode, new_size);
1682 	}
1683 
1684 	return err;
1685 }
1686 
1687 static long f2fs_fallocate(struct file *file, int mode,
1688 				loff_t offset, loff_t len)
1689 {
1690 	struct inode *inode = file_inode(file);
1691 	long ret = 0;
1692 
1693 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode))))
1694 		return -EIO;
1695 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(inode)))
1696 		return -ENOSPC;
1697 	if (!f2fs_is_compress_backend_ready(inode))
1698 		return -EOPNOTSUPP;
1699 
1700 	/* f2fs only support ->fallocate for regular file */
1701 	if (!S_ISREG(inode->i_mode))
1702 		return -EINVAL;
1703 
1704 	if (IS_ENCRYPTED(inode) &&
1705 		(mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
1706 		return -EOPNOTSUPP;
1707 
1708 	if (f2fs_compressed_file(inode) &&
1709 		(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE |
1710 			FALLOC_FL_ZERO_RANGE | FALLOC_FL_INSERT_RANGE)))
1711 		return -EOPNOTSUPP;
1712 
1713 	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
1714 			FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
1715 			FALLOC_FL_INSERT_RANGE))
1716 		return -EOPNOTSUPP;
1717 
1718 	inode_lock(inode);
1719 
1720 	if (mode & FALLOC_FL_PUNCH_HOLE) {
1721 		if (offset >= inode->i_size)
1722 			goto out;
1723 
1724 		ret = punch_hole(inode, offset, len);
1725 	} else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
1726 		ret = f2fs_collapse_range(inode, offset, len);
1727 	} else if (mode & FALLOC_FL_ZERO_RANGE) {
1728 		ret = f2fs_zero_range(inode, offset, len, mode);
1729 	} else if (mode & FALLOC_FL_INSERT_RANGE) {
1730 		ret = f2fs_insert_range(inode, offset, len);
1731 	} else {
1732 		ret = expand_inode_data(inode, offset, len, mode);
1733 	}
1734 
1735 	if (!ret) {
1736 		inode->i_mtime = inode->i_ctime = current_time(inode);
1737 		f2fs_mark_inode_dirty_sync(inode, false);
1738 		f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
1739 	}
1740 
1741 out:
1742 	inode_unlock(inode);
1743 
1744 	trace_f2fs_fallocate(inode, mode, offset, len, ret);
1745 	return ret;
1746 }
1747 
1748 static int f2fs_release_file(struct inode *inode, struct file *filp)
1749 {
1750 	/*
1751 	 * f2fs_relase_file is called at every close calls. So we should
1752 	 * not drop any inmemory pages by close called by other process.
1753 	 */
1754 	if (!(filp->f_mode & FMODE_WRITE) ||
1755 			atomic_read(&inode->i_writecount) != 1)
1756 		return 0;
1757 
1758 	/* some remained atomic pages should discarded */
1759 	if (f2fs_is_atomic_file(inode))
1760 		f2fs_drop_inmem_pages(inode);
1761 	if (f2fs_is_volatile_file(inode)) {
1762 		set_inode_flag(inode, FI_DROP_CACHE);
1763 		filemap_fdatawrite(inode->i_mapping);
1764 		clear_inode_flag(inode, FI_DROP_CACHE);
1765 		clear_inode_flag(inode, FI_VOLATILE_FILE);
1766 		stat_dec_volatile_write(inode);
1767 	}
1768 	return 0;
1769 }
1770 
1771 static int f2fs_file_flush(struct file *file, fl_owner_t id)
1772 {
1773 	struct inode *inode = file_inode(file);
1774 
1775 	/*
1776 	 * If the process doing a transaction is crashed, we should do
1777 	 * roll-back. Otherwise, other reader/write can see corrupted database
1778 	 * until all the writers close its file. Since this should be done
1779 	 * before dropping file lock, it needs to do in ->flush.
1780 	 */
1781 	if (f2fs_is_atomic_file(inode) &&
1782 			F2FS_I(inode)->inmem_task == current)
1783 		f2fs_drop_inmem_pages(inode);
1784 	return 0;
1785 }
1786 
1787 static int f2fs_setflags_common(struct inode *inode, u32 iflags, u32 mask)
1788 {
1789 	struct f2fs_inode_info *fi = F2FS_I(inode);
1790 
1791 	/* Is it quota file? Do not allow user to mess with it */
1792 	if (IS_NOQUOTA(inode))
1793 		return -EPERM;
1794 
1795 	if ((iflags ^ fi->i_flags) & F2FS_CASEFOLD_FL) {
1796 		if (!f2fs_sb_has_casefold(F2FS_I_SB(inode)))
1797 			return -EOPNOTSUPP;
1798 		if (!f2fs_empty_dir(inode))
1799 			return -ENOTEMPTY;
1800 	}
1801 
1802 	if (iflags & (F2FS_COMPR_FL | F2FS_NOCOMP_FL)) {
1803 		if (!f2fs_sb_has_compression(F2FS_I_SB(inode)))
1804 			return -EOPNOTSUPP;
1805 		if ((iflags & F2FS_COMPR_FL) && (iflags & F2FS_NOCOMP_FL))
1806 			return -EINVAL;
1807 	}
1808 
1809 	if ((iflags ^ fi->i_flags) & F2FS_COMPR_FL) {
1810 		if (S_ISREG(inode->i_mode) &&
1811 			(fi->i_flags & F2FS_COMPR_FL || i_size_read(inode) ||
1812 						F2FS_HAS_BLOCKS(inode)))
1813 			return -EINVAL;
1814 		if (iflags & F2FS_NOCOMP_FL)
1815 			return -EINVAL;
1816 		if (iflags & F2FS_COMPR_FL) {
1817 			int err = f2fs_convert_inline_inode(inode);
1818 
1819 			if (err)
1820 				return err;
1821 
1822 			if (!f2fs_may_compress(inode))
1823 				return -EINVAL;
1824 
1825 			set_compress_context(inode);
1826 		}
1827 	}
1828 	if ((iflags ^ fi->i_flags) & F2FS_NOCOMP_FL) {
1829 		if (fi->i_flags & F2FS_COMPR_FL)
1830 			return -EINVAL;
1831 	}
1832 
1833 	fi->i_flags = iflags | (fi->i_flags & ~mask);
1834 	f2fs_bug_on(F2FS_I_SB(inode), (fi->i_flags & F2FS_COMPR_FL) &&
1835 					(fi->i_flags & F2FS_NOCOMP_FL));
1836 
1837 	if (fi->i_flags & F2FS_PROJINHERIT_FL)
1838 		set_inode_flag(inode, FI_PROJ_INHERIT);
1839 	else
1840 		clear_inode_flag(inode, FI_PROJ_INHERIT);
1841 
1842 	inode->i_ctime = current_time(inode);
1843 	f2fs_set_inode_flags(inode);
1844 	f2fs_mark_inode_dirty_sync(inode, true);
1845 	return 0;
1846 }
1847 
1848 /* FS_IOC_GETFLAGS and FS_IOC_SETFLAGS support */
1849 
1850 /*
1851  * To make a new on-disk f2fs i_flag gettable via FS_IOC_GETFLAGS, add an entry
1852  * for it to f2fs_fsflags_map[], and add its FS_*_FL equivalent to
1853  * F2FS_GETTABLE_FS_FL.  To also make it settable via FS_IOC_SETFLAGS, also add
1854  * its FS_*_FL equivalent to F2FS_SETTABLE_FS_FL.
1855  */
1856 
1857 static const struct {
1858 	u32 iflag;
1859 	u32 fsflag;
1860 } f2fs_fsflags_map[] = {
1861 	{ F2FS_COMPR_FL,	FS_COMPR_FL },
1862 	{ F2FS_SYNC_FL,		FS_SYNC_FL },
1863 	{ F2FS_IMMUTABLE_FL,	FS_IMMUTABLE_FL },
1864 	{ F2FS_APPEND_FL,	FS_APPEND_FL },
1865 	{ F2FS_NODUMP_FL,	FS_NODUMP_FL },
1866 	{ F2FS_NOATIME_FL,	FS_NOATIME_FL },
1867 	{ F2FS_NOCOMP_FL,	FS_NOCOMP_FL },
1868 	{ F2FS_INDEX_FL,	FS_INDEX_FL },
1869 	{ F2FS_DIRSYNC_FL,	FS_DIRSYNC_FL },
1870 	{ F2FS_PROJINHERIT_FL,	FS_PROJINHERIT_FL },
1871 	{ F2FS_CASEFOLD_FL,	FS_CASEFOLD_FL },
1872 };
1873 
1874 #define F2FS_GETTABLE_FS_FL (		\
1875 		FS_COMPR_FL |		\
1876 		FS_SYNC_FL |		\
1877 		FS_IMMUTABLE_FL |	\
1878 		FS_APPEND_FL |		\
1879 		FS_NODUMP_FL |		\
1880 		FS_NOATIME_FL |		\
1881 		FS_NOCOMP_FL |		\
1882 		FS_INDEX_FL |		\
1883 		FS_DIRSYNC_FL |		\
1884 		FS_PROJINHERIT_FL |	\
1885 		FS_ENCRYPT_FL |		\
1886 		FS_INLINE_DATA_FL |	\
1887 		FS_NOCOW_FL |		\
1888 		FS_VERITY_FL |		\
1889 		FS_CASEFOLD_FL)
1890 
1891 #define F2FS_SETTABLE_FS_FL (		\
1892 		FS_COMPR_FL |		\
1893 		FS_SYNC_FL |		\
1894 		FS_IMMUTABLE_FL |	\
1895 		FS_APPEND_FL |		\
1896 		FS_NODUMP_FL |		\
1897 		FS_NOATIME_FL |		\
1898 		FS_NOCOMP_FL |		\
1899 		FS_DIRSYNC_FL |		\
1900 		FS_PROJINHERIT_FL |	\
1901 		FS_CASEFOLD_FL)
1902 
1903 /* Convert f2fs on-disk i_flags to FS_IOC_{GET,SET}FLAGS flags */
1904 static inline u32 f2fs_iflags_to_fsflags(u32 iflags)
1905 {
1906 	u32 fsflags = 0;
1907 	int i;
1908 
1909 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1910 		if (iflags & f2fs_fsflags_map[i].iflag)
1911 			fsflags |= f2fs_fsflags_map[i].fsflag;
1912 
1913 	return fsflags;
1914 }
1915 
1916 /* Convert FS_IOC_{GET,SET}FLAGS flags to f2fs on-disk i_flags */
1917 static inline u32 f2fs_fsflags_to_iflags(u32 fsflags)
1918 {
1919 	u32 iflags = 0;
1920 	int i;
1921 
1922 	for (i = 0; i < ARRAY_SIZE(f2fs_fsflags_map); i++)
1923 		if (fsflags & f2fs_fsflags_map[i].fsflag)
1924 			iflags |= f2fs_fsflags_map[i].iflag;
1925 
1926 	return iflags;
1927 }
1928 
1929 static int f2fs_ioc_getflags(struct file *filp, unsigned long arg)
1930 {
1931 	struct inode *inode = file_inode(filp);
1932 	struct f2fs_inode_info *fi = F2FS_I(inode);
1933 	u32 fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1934 
1935 	if (IS_ENCRYPTED(inode))
1936 		fsflags |= FS_ENCRYPT_FL;
1937 	if (IS_VERITY(inode))
1938 		fsflags |= FS_VERITY_FL;
1939 	if (f2fs_has_inline_data(inode) || f2fs_has_inline_dentry(inode))
1940 		fsflags |= FS_INLINE_DATA_FL;
1941 	if (is_inode_flag_set(inode, FI_PIN_FILE))
1942 		fsflags |= FS_NOCOW_FL;
1943 
1944 	fsflags &= F2FS_GETTABLE_FS_FL;
1945 
1946 	return put_user(fsflags, (int __user *)arg);
1947 }
1948 
1949 static int f2fs_ioc_setflags(struct file *filp, unsigned long arg)
1950 {
1951 	struct inode *inode = file_inode(filp);
1952 	struct f2fs_inode_info *fi = F2FS_I(inode);
1953 	u32 fsflags, old_fsflags;
1954 	u32 iflags;
1955 	int ret;
1956 
1957 	if (!inode_owner_or_capable(inode))
1958 		return -EACCES;
1959 
1960 	if (get_user(fsflags, (int __user *)arg))
1961 		return -EFAULT;
1962 
1963 	if (fsflags & ~F2FS_GETTABLE_FS_FL)
1964 		return -EOPNOTSUPP;
1965 	fsflags &= F2FS_SETTABLE_FS_FL;
1966 
1967 	iflags = f2fs_fsflags_to_iflags(fsflags);
1968 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
1969 		return -EOPNOTSUPP;
1970 
1971 	ret = mnt_want_write_file(filp);
1972 	if (ret)
1973 		return ret;
1974 
1975 	inode_lock(inode);
1976 
1977 	old_fsflags = f2fs_iflags_to_fsflags(fi->i_flags);
1978 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
1979 	if (ret)
1980 		goto out;
1981 
1982 	ret = f2fs_setflags_common(inode, iflags,
1983 			f2fs_fsflags_to_iflags(F2FS_SETTABLE_FS_FL));
1984 out:
1985 	inode_unlock(inode);
1986 	mnt_drop_write_file(filp);
1987 	return ret;
1988 }
1989 
1990 static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
1991 {
1992 	struct inode *inode = file_inode(filp);
1993 
1994 	return put_user(inode->i_generation, (int __user *)arg);
1995 }
1996 
1997 static int f2fs_ioc_start_atomic_write(struct file *filp)
1998 {
1999 	struct inode *inode = file_inode(filp);
2000 	struct f2fs_inode_info *fi = F2FS_I(inode);
2001 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2002 	int ret;
2003 
2004 	if (!inode_owner_or_capable(inode))
2005 		return -EACCES;
2006 
2007 	if (!S_ISREG(inode->i_mode))
2008 		return -EINVAL;
2009 
2010 	if (filp->f_flags & O_DIRECT)
2011 		return -EINVAL;
2012 
2013 	ret = mnt_want_write_file(filp);
2014 	if (ret)
2015 		return ret;
2016 
2017 	inode_lock(inode);
2018 
2019 	f2fs_disable_compressed_file(inode);
2020 
2021 	if (f2fs_is_atomic_file(inode)) {
2022 		if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST))
2023 			ret = -EINVAL;
2024 		goto out;
2025 	}
2026 
2027 	ret = f2fs_convert_inline_inode(inode);
2028 	if (ret)
2029 		goto out;
2030 
2031 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2032 
2033 	/*
2034 	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
2035 	 * f2fs_is_atomic_file.
2036 	 */
2037 	if (get_dirty_pages(inode))
2038 		f2fs_warn(F2FS_I_SB(inode), "Unexpected flush for atomic writes: ino=%lu, npages=%u",
2039 			  inode->i_ino, get_dirty_pages(inode));
2040 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
2041 	if (ret) {
2042 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2043 		goto out;
2044 	}
2045 
2046 	spin_lock(&sbi->inode_lock[ATOMIC_FILE]);
2047 	if (list_empty(&fi->inmem_ilist))
2048 		list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
2049 	sbi->atomic_files++;
2050 	spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
2051 
2052 	/* add inode in inmem_list first and set atomic_file */
2053 	set_inode_flag(inode, FI_ATOMIC_FILE);
2054 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2055 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
2056 
2057 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2058 	F2FS_I(inode)->inmem_task = current;
2059 	stat_update_max_atomic_write(inode);
2060 out:
2061 	inode_unlock(inode);
2062 	mnt_drop_write_file(filp);
2063 	return ret;
2064 }
2065 
2066 static int f2fs_ioc_commit_atomic_write(struct file *filp)
2067 {
2068 	struct inode *inode = file_inode(filp);
2069 	int ret;
2070 
2071 	if (!inode_owner_or_capable(inode))
2072 		return -EACCES;
2073 
2074 	ret = mnt_want_write_file(filp);
2075 	if (ret)
2076 		return ret;
2077 
2078 	f2fs_balance_fs(F2FS_I_SB(inode), true);
2079 
2080 	inode_lock(inode);
2081 
2082 	if (f2fs_is_volatile_file(inode)) {
2083 		ret = -EINVAL;
2084 		goto err_out;
2085 	}
2086 
2087 	if (f2fs_is_atomic_file(inode)) {
2088 		ret = f2fs_commit_inmem_pages(inode);
2089 		if (ret)
2090 			goto err_out;
2091 
2092 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2093 		if (!ret)
2094 			f2fs_drop_inmem_pages(inode);
2095 	} else {
2096 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 1, false);
2097 	}
2098 err_out:
2099 	if (is_inode_flag_set(inode, FI_ATOMIC_REVOKE_REQUEST)) {
2100 		clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2101 		ret = -EINVAL;
2102 	}
2103 	inode_unlock(inode);
2104 	mnt_drop_write_file(filp);
2105 	return ret;
2106 }
2107 
2108 static int f2fs_ioc_start_volatile_write(struct file *filp)
2109 {
2110 	struct inode *inode = file_inode(filp);
2111 	int ret;
2112 
2113 	if (!inode_owner_or_capable(inode))
2114 		return -EACCES;
2115 
2116 	if (!S_ISREG(inode->i_mode))
2117 		return -EINVAL;
2118 
2119 	ret = mnt_want_write_file(filp);
2120 	if (ret)
2121 		return ret;
2122 
2123 	inode_lock(inode);
2124 
2125 	if (f2fs_is_volatile_file(inode))
2126 		goto out;
2127 
2128 	ret = f2fs_convert_inline_inode(inode);
2129 	if (ret)
2130 		goto out;
2131 
2132 	stat_inc_volatile_write(inode);
2133 	stat_update_max_volatile_write(inode);
2134 
2135 	set_inode_flag(inode, FI_VOLATILE_FILE);
2136 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2137 out:
2138 	inode_unlock(inode);
2139 	mnt_drop_write_file(filp);
2140 	return ret;
2141 }
2142 
2143 static int f2fs_ioc_release_volatile_write(struct file *filp)
2144 {
2145 	struct inode *inode = file_inode(filp);
2146 	int ret;
2147 
2148 	if (!inode_owner_or_capable(inode))
2149 		return -EACCES;
2150 
2151 	ret = mnt_want_write_file(filp);
2152 	if (ret)
2153 		return ret;
2154 
2155 	inode_lock(inode);
2156 
2157 	if (!f2fs_is_volatile_file(inode))
2158 		goto out;
2159 
2160 	if (!f2fs_is_first_block_written(inode)) {
2161 		ret = truncate_partial_data_page(inode, 0, true);
2162 		goto out;
2163 	}
2164 
2165 	ret = punch_hole(inode, 0, F2FS_BLKSIZE);
2166 out:
2167 	inode_unlock(inode);
2168 	mnt_drop_write_file(filp);
2169 	return ret;
2170 }
2171 
2172 static int f2fs_ioc_abort_volatile_write(struct file *filp)
2173 {
2174 	struct inode *inode = file_inode(filp);
2175 	int ret;
2176 
2177 	if (!inode_owner_or_capable(inode))
2178 		return -EACCES;
2179 
2180 	ret = mnt_want_write_file(filp);
2181 	if (ret)
2182 		return ret;
2183 
2184 	inode_lock(inode);
2185 
2186 	if (f2fs_is_atomic_file(inode))
2187 		f2fs_drop_inmem_pages(inode);
2188 	if (f2fs_is_volatile_file(inode)) {
2189 		clear_inode_flag(inode, FI_VOLATILE_FILE);
2190 		stat_dec_volatile_write(inode);
2191 		ret = f2fs_do_sync_file(filp, 0, LLONG_MAX, 0, true);
2192 	}
2193 
2194 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
2195 
2196 	inode_unlock(inode);
2197 
2198 	mnt_drop_write_file(filp);
2199 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2200 	return ret;
2201 }
2202 
2203 static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
2204 {
2205 	struct inode *inode = file_inode(filp);
2206 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2207 	struct super_block *sb = sbi->sb;
2208 	__u32 in;
2209 	int ret = 0;
2210 
2211 	if (!capable(CAP_SYS_ADMIN))
2212 		return -EPERM;
2213 
2214 	if (get_user(in, (__u32 __user *)arg))
2215 		return -EFAULT;
2216 
2217 	if (in != F2FS_GOING_DOWN_FULLSYNC) {
2218 		ret = mnt_want_write_file(filp);
2219 		if (ret)
2220 			return ret;
2221 	}
2222 
2223 	switch (in) {
2224 	case F2FS_GOING_DOWN_FULLSYNC:
2225 		sb = freeze_bdev(sb->s_bdev);
2226 		if (IS_ERR(sb)) {
2227 			ret = PTR_ERR(sb);
2228 			goto out;
2229 		}
2230 		if (sb) {
2231 			f2fs_stop_checkpoint(sbi, false);
2232 			set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2233 			thaw_bdev(sb->s_bdev, sb);
2234 		}
2235 		break;
2236 	case F2FS_GOING_DOWN_METASYNC:
2237 		/* do checkpoint only */
2238 		ret = f2fs_sync_fs(sb, 1);
2239 		if (ret)
2240 			goto out;
2241 		f2fs_stop_checkpoint(sbi, false);
2242 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2243 		break;
2244 	case F2FS_GOING_DOWN_NOSYNC:
2245 		f2fs_stop_checkpoint(sbi, false);
2246 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2247 		break;
2248 	case F2FS_GOING_DOWN_METAFLUSH:
2249 		f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_META_IO);
2250 		f2fs_stop_checkpoint(sbi, false);
2251 		set_sbi_flag(sbi, SBI_IS_SHUTDOWN);
2252 		break;
2253 	case F2FS_GOING_DOWN_NEED_FSCK:
2254 		set_sbi_flag(sbi, SBI_NEED_FSCK);
2255 		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
2256 		set_sbi_flag(sbi, SBI_IS_DIRTY);
2257 		/* do checkpoint only */
2258 		ret = f2fs_sync_fs(sb, 1);
2259 		goto out;
2260 	default:
2261 		ret = -EINVAL;
2262 		goto out;
2263 	}
2264 
2265 	f2fs_stop_gc_thread(sbi);
2266 	f2fs_stop_discard_thread(sbi);
2267 
2268 	f2fs_drop_discard_cmd(sbi);
2269 	clear_opt(sbi, DISCARD);
2270 
2271 	f2fs_update_time(sbi, REQ_TIME);
2272 out:
2273 	if (in != F2FS_GOING_DOWN_FULLSYNC)
2274 		mnt_drop_write_file(filp);
2275 
2276 	trace_f2fs_shutdown(sbi, in, ret);
2277 
2278 	return ret;
2279 }
2280 
2281 static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
2282 {
2283 	struct inode *inode = file_inode(filp);
2284 	struct super_block *sb = inode->i_sb;
2285 	struct request_queue *q = bdev_get_queue(sb->s_bdev);
2286 	struct fstrim_range range;
2287 	int ret;
2288 
2289 	if (!capable(CAP_SYS_ADMIN))
2290 		return -EPERM;
2291 
2292 	if (!f2fs_hw_support_discard(F2FS_SB(sb)))
2293 		return -EOPNOTSUPP;
2294 
2295 	if (copy_from_user(&range, (struct fstrim_range __user *)arg,
2296 				sizeof(range)))
2297 		return -EFAULT;
2298 
2299 	ret = mnt_want_write_file(filp);
2300 	if (ret)
2301 		return ret;
2302 
2303 	range.minlen = max((unsigned int)range.minlen,
2304 				q->limits.discard_granularity);
2305 	ret = f2fs_trim_fs(F2FS_SB(sb), &range);
2306 	mnt_drop_write_file(filp);
2307 	if (ret < 0)
2308 		return ret;
2309 
2310 	if (copy_to_user((struct fstrim_range __user *)arg, &range,
2311 				sizeof(range)))
2312 		return -EFAULT;
2313 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2314 	return 0;
2315 }
2316 
2317 static bool uuid_is_nonzero(__u8 u[16])
2318 {
2319 	int i;
2320 
2321 	for (i = 0; i < 16; i++)
2322 		if (u[i])
2323 			return true;
2324 	return false;
2325 }
2326 
2327 static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
2328 {
2329 	struct inode *inode = file_inode(filp);
2330 
2331 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(inode)))
2332 		return -EOPNOTSUPP;
2333 
2334 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
2335 
2336 	return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
2337 }
2338 
2339 static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
2340 {
2341 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2342 		return -EOPNOTSUPP;
2343 	return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
2344 }
2345 
2346 static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
2347 {
2348 	struct inode *inode = file_inode(filp);
2349 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2350 	int err;
2351 
2352 	if (!f2fs_sb_has_encrypt(sbi))
2353 		return -EOPNOTSUPP;
2354 
2355 	err = mnt_want_write_file(filp);
2356 	if (err)
2357 		return err;
2358 
2359 	down_write(&sbi->sb_lock);
2360 
2361 	if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
2362 		goto got_it;
2363 
2364 	/* update superblock with uuid */
2365 	generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
2366 
2367 	err = f2fs_commit_super(sbi, false);
2368 	if (err) {
2369 		/* undo new data */
2370 		memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
2371 		goto out_err;
2372 	}
2373 got_it:
2374 	if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
2375 									16))
2376 		err = -EFAULT;
2377 out_err:
2378 	up_write(&sbi->sb_lock);
2379 	mnt_drop_write_file(filp);
2380 	return err;
2381 }
2382 
2383 static int f2fs_ioc_get_encryption_policy_ex(struct file *filp,
2384 					     unsigned long arg)
2385 {
2386 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2387 		return -EOPNOTSUPP;
2388 
2389 	return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
2390 }
2391 
2392 static int f2fs_ioc_add_encryption_key(struct file *filp, unsigned long arg)
2393 {
2394 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2395 		return -EOPNOTSUPP;
2396 
2397 	return fscrypt_ioctl_add_key(filp, (void __user *)arg);
2398 }
2399 
2400 static int f2fs_ioc_remove_encryption_key(struct file *filp, unsigned long arg)
2401 {
2402 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2403 		return -EOPNOTSUPP;
2404 
2405 	return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
2406 }
2407 
2408 static int f2fs_ioc_remove_encryption_key_all_users(struct file *filp,
2409 						    unsigned long arg)
2410 {
2411 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2412 		return -EOPNOTSUPP;
2413 
2414 	return fscrypt_ioctl_remove_key_all_users(filp, (void __user *)arg);
2415 }
2416 
2417 static int f2fs_ioc_get_encryption_key_status(struct file *filp,
2418 					      unsigned long arg)
2419 {
2420 	if (!f2fs_sb_has_encrypt(F2FS_I_SB(file_inode(filp))))
2421 		return -EOPNOTSUPP;
2422 
2423 	return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
2424 }
2425 
2426 static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
2427 {
2428 	struct inode *inode = file_inode(filp);
2429 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2430 	__u32 sync;
2431 	int ret;
2432 
2433 	if (!capable(CAP_SYS_ADMIN))
2434 		return -EPERM;
2435 
2436 	if (get_user(sync, (__u32 __user *)arg))
2437 		return -EFAULT;
2438 
2439 	if (f2fs_readonly(sbi->sb))
2440 		return -EROFS;
2441 
2442 	ret = mnt_want_write_file(filp);
2443 	if (ret)
2444 		return ret;
2445 
2446 	if (!sync) {
2447 		if (!down_write_trylock(&sbi->gc_lock)) {
2448 			ret = -EBUSY;
2449 			goto out;
2450 		}
2451 	} else {
2452 		down_write(&sbi->gc_lock);
2453 	}
2454 
2455 	ret = f2fs_gc(sbi, sync, true, NULL_SEGNO);
2456 out:
2457 	mnt_drop_write_file(filp);
2458 	return ret;
2459 }
2460 
2461 static int f2fs_ioc_gc_range(struct file *filp, unsigned long arg)
2462 {
2463 	struct inode *inode = file_inode(filp);
2464 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2465 	struct f2fs_gc_range range;
2466 	u64 end;
2467 	int ret;
2468 
2469 	if (!capable(CAP_SYS_ADMIN))
2470 		return -EPERM;
2471 
2472 	if (copy_from_user(&range, (struct f2fs_gc_range __user *)arg,
2473 							sizeof(range)))
2474 		return -EFAULT;
2475 
2476 	if (f2fs_readonly(sbi->sb))
2477 		return -EROFS;
2478 
2479 	end = range.start + range.len;
2480 	if (end < range.start || range.start < MAIN_BLKADDR(sbi) ||
2481 					end >= MAX_BLKADDR(sbi))
2482 		return -EINVAL;
2483 
2484 	ret = mnt_want_write_file(filp);
2485 	if (ret)
2486 		return ret;
2487 
2488 do_more:
2489 	if (!range.sync) {
2490 		if (!down_write_trylock(&sbi->gc_lock)) {
2491 			ret = -EBUSY;
2492 			goto out;
2493 		}
2494 	} else {
2495 		down_write(&sbi->gc_lock);
2496 	}
2497 
2498 	ret = f2fs_gc(sbi, range.sync, true, GET_SEGNO(sbi, range.start));
2499 	range.start += BLKS_PER_SEC(sbi);
2500 	if (range.start <= end)
2501 		goto do_more;
2502 out:
2503 	mnt_drop_write_file(filp);
2504 	return ret;
2505 }
2506 
2507 static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
2508 {
2509 	struct inode *inode = file_inode(filp);
2510 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2511 	int ret;
2512 
2513 	if (!capable(CAP_SYS_ADMIN))
2514 		return -EPERM;
2515 
2516 	if (f2fs_readonly(sbi->sb))
2517 		return -EROFS;
2518 
2519 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
2520 		f2fs_info(sbi, "Skipping Checkpoint. Checkpoints currently disabled.");
2521 		return -EINVAL;
2522 	}
2523 
2524 	ret = mnt_want_write_file(filp);
2525 	if (ret)
2526 		return ret;
2527 
2528 	ret = f2fs_sync_fs(sbi->sb, 1);
2529 
2530 	mnt_drop_write_file(filp);
2531 	return ret;
2532 }
2533 
2534 static int f2fs_defragment_range(struct f2fs_sb_info *sbi,
2535 					struct file *filp,
2536 					struct f2fs_defragment *range)
2537 {
2538 	struct inode *inode = file_inode(filp);
2539 	struct f2fs_map_blocks map = { .m_next_extent = NULL,
2540 					.m_seg_type = NO_CHECK_TYPE ,
2541 					.m_may_create = false };
2542 	struct extent_info ei = {0, 0, 0};
2543 	pgoff_t pg_start, pg_end, next_pgofs;
2544 	unsigned int blk_per_seg = sbi->blocks_per_seg;
2545 	unsigned int total = 0, sec_num;
2546 	block_t blk_end = 0;
2547 	bool fragmented = false;
2548 	int err;
2549 
2550 	/* if in-place-update policy is enabled, don't waste time here */
2551 	if (f2fs_should_update_inplace(inode, NULL))
2552 		return -EINVAL;
2553 
2554 	pg_start = range->start >> PAGE_SHIFT;
2555 	pg_end = (range->start + range->len) >> PAGE_SHIFT;
2556 
2557 	f2fs_balance_fs(sbi, true);
2558 
2559 	inode_lock(inode);
2560 
2561 	/* writeback all dirty pages in the range */
2562 	err = filemap_write_and_wait_range(inode->i_mapping, range->start,
2563 						range->start + range->len - 1);
2564 	if (err)
2565 		goto out;
2566 
2567 	/*
2568 	 * lookup mapping info in extent cache, skip defragmenting if physical
2569 	 * block addresses are continuous.
2570 	 */
2571 	if (f2fs_lookup_extent_cache(inode, pg_start, &ei)) {
2572 		if (ei.fofs + ei.len >= pg_end)
2573 			goto out;
2574 	}
2575 
2576 	map.m_lblk = pg_start;
2577 	map.m_next_pgofs = &next_pgofs;
2578 
2579 	/*
2580 	 * lookup mapping info in dnode page cache, skip defragmenting if all
2581 	 * physical block addresses are continuous even if there are hole(s)
2582 	 * in logical blocks.
2583 	 */
2584 	while (map.m_lblk < pg_end) {
2585 		map.m_len = pg_end - map.m_lblk;
2586 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2587 		if (err)
2588 			goto out;
2589 
2590 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2591 			map.m_lblk = next_pgofs;
2592 			continue;
2593 		}
2594 
2595 		if (blk_end && blk_end != map.m_pblk)
2596 			fragmented = true;
2597 
2598 		/* record total count of block that we're going to move */
2599 		total += map.m_len;
2600 
2601 		blk_end = map.m_pblk + map.m_len;
2602 
2603 		map.m_lblk += map.m_len;
2604 	}
2605 
2606 	if (!fragmented) {
2607 		total = 0;
2608 		goto out;
2609 	}
2610 
2611 	sec_num = DIV_ROUND_UP(total, BLKS_PER_SEC(sbi));
2612 
2613 	/*
2614 	 * make sure there are enough free section for LFS allocation, this can
2615 	 * avoid defragment running in SSR mode when free section are allocated
2616 	 * intensively
2617 	 */
2618 	if (has_not_enough_free_secs(sbi, 0, sec_num)) {
2619 		err = -EAGAIN;
2620 		goto out;
2621 	}
2622 
2623 	map.m_lblk = pg_start;
2624 	map.m_len = pg_end - pg_start;
2625 	total = 0;
2626 
2627 	while (map.m_lblk < pg_end) {
2628 		pgoff_t idx;
2629 		int cnt = 0;
2630 
2631 do_map:
2632 		map.m_len = pg_end - map.m_lblk;
2633 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_DEFAULT);
2634 		if (err)
2635 			goto clear_out;
2636 
2637 		if (!(map.m_flags & F2FS_MAP_FLAGS)) {
2638 			map.m_lblk = next_pgofs;
2639 			goto check;
2640 		}
2641 
2642 		set_inode_flag(inode, FI_DO_DEFRAG);
2643 
2644 		idx = map.m_lblk;
2645 		while (idx < map.m_lblk + map.m_len && cnt < blk_per_seg) {
2646 			struct page *page;
2647 
2648 			page = f2fs_get_lock_data_page(inode, idx, true);
2649 			if (IS_ERR(page)) {
2650 				err = PTR_ERR(page);
2651 				goto clear_out;
2652 			}
2653 
2654 			set_page_dirty(page);
2655 			f2fs_put_page(page, 1);
2656 
2657 			idx++;
2658 			cnt++;
2659 			total++;
2660 		}
2661 
2662 		map.m_lblk = idx;
2663 check:
2664 		if (map.m_lblk < pg_end && cnt < blk_per_seg)
2665 			goto do_map;
2666 
2667 		clear_inode_flag(inode, FI_DO_DEFRAG);
2668 
2669 		err = filemap_fdatawrite(inode->i_mapping);
2670 		if (err)
2671 			goto out;
2672 	}
2673 clear_out:
2674 	clear_inode_flag(inode, FI_DO_DEFRAG);
2675 out:
2676 	inode_unlock(inode);
2677 	if (!err)
2678 		range->len = (u64)total << PAGE_SHIFT;
2679 	return err;
2680 }
2681 
2682 static int f2fs_ioc_defragment(struct file *filp, unsigned long arg)
2683 {
2684 	struct inode *inode = file_inode(filp);
2685 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2686 	struct f2fs_defragment range;
2687 	int err;
2688 
2689 	if (!capable(CAP_SYS_ADMIN))
2690 		return -EPERM;
2691 
2692 	if (!S_ISREG(inode->i_mode) || f2fs_is_atomic_file(inode))
2693 		return -EINVAL;
2694 
2695 	if (f2fs_readonly(sbi->sb))
2696 		return -EROFS;
2697 
2698 	if (copy_from_user(&range, (struct f2fs_defragment __user *)arg,
2699 							sizeof(range)))
2700 		return -EFAULT;
2701 
2702 	/* verify alignment of offset & size */
2703 	if (range.start & (F2FS_BLKSIZE - 1) || range.len & (F2FS_BLKSIZE - 1))
2704 		return -EINVAL;
2705 
2706 	if (unlikely((range.start + range.len) >> PAGE_SHIFT >
2707 					sbi->max_file_blocks))
2708 		return -EINVAL;
2709 
2710 	err = mnt_want_write_file(filp);
2711 	if (err)
2712 		return err;
2713 
2714 	err = f2fs_defragment_range(sbi, filp, &range);
2715 	mnt_drop_write_file(filp);
2716 
2717 	f2fs_update_time(sbi, REQ_TIME);
2718 	if (err < 0)
2719 		return err;
2720 
2721 	if (copy_to_user((struct f2fs_defragment __user *)arg, &range,
2722 							sizeof(range)))
2723 		return -EFAULT;
2724 
2725 	return 0;
2726 }
2727 
2728 static int f2fs_move_file_range(struct file *file_in, loff_t pos_in,
2729 			struct file *file_out, loff_t pos_out, size_t len)
2730 {
2731 	struct inode *src = file_inode(file_in);
2732 	struct inode *dst = file_inode(file_out);
2733 	struct f2fs_sb_info *sbi = F2FS_I_SB(src);
2734 	size_t olen = len, dst_max_i_size = 0;
2735 	size_t dst_osize;
2736 	int ret;
2737 
2738 	if (file_in->f_path.mnt != file_out->f_path.mnt ||
2739 				src->i_sb != dst->i_sb)
2740 		return -EXDEV;
2741 
2742 	if (unlikely(f2fs_readonly(src->i_sb)))
2743 		return -EROFS;
2744 
2745 	if (!S_ISREG(src->i_mode) || !S_ISREG(dst->i_mode))
2746 		return -EINVAL;
2747 
2748 	if (IS_ENCRYPTED(src) || IS_ENCRYPTED(dst))
2749 		return -EOPNOTSUPP;
2750 
2751 	if (src == dst) {
2752 		if (pos_in == pos_out)
2753 			return 0;
2754 		if (pos_out > pos_in && pos_out < pos_in + len)
2755 			return -EINVAL;
2756 	}
2757 
2758 	inode_lock(src);
2759 	if (src != dst) {
2760 		ret = -EBUSY;
2761 		if (!inode_trylock(dst))
2762 			goto out;
2763 	}
2764 
2765 	ret = -EINVAL;
2766 	if (pos_in + len > src->i_size || pos_in + len < pos_in)
2767 		goto out_unlock;
2768 	if (len == 0)
2769 		olen = len = src->i_size - pos_in;
2770 	if (pos_in + len == src->i_size)
2771 		len = ALIGN(src->i_size, F2FS_BLKSIZE) - pos_in;
2772 	if (len == 0) {
2773 		ret = 0;
2774 		goto out_unlock;
2775 	}
2776 
2777 	dst_osize = dst->i_size;
2778 	if (pos_out + olen > dst->i_size)
2779 		dst_max_i_size = pos_out + olen;
2780 
2781 	/* verify the end result is block aligned */
2782 	if (!IS_ALIGNED(pos_in, F2FS_BLKSIZE) ||
2783 			!IS_ALIGNED(pos_in + len, F2FS_BLKSIZE) ||
2784 			!IS_ALIGNED(pos_out, F2FS_BLKSIZE))
2785 		goto out_unlock;
2786 
2787 	ret = f2fs_convert_inline_inode(src);
2788 	if (ret)
2789 		goto out_unlock;
2790 
2791 	ret = f2fs_convert_inline_inode(dst);
2792 	if (ret)
2793 		goto out_unlock;
2794 
2795 	/* write out all dirty pages from offset */
2796 	ret = filemap_write_and_wait_range(src->i_mapping,
2797 					pos_in, pos_in + len);
2798 	if (ret)
2799 		goto out_unlock;
2800 
2801 	ret = filemap_write_and_wait_range(dst->i_mapping,
2802 					pos_out, pos_out + len);
2803 	if (ret)
2804 		goto out_unlock;
2805 
2806 	f2fs_balance_fs(sbi, true);
2807 
2808 	down_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2809 	if (src != dst) {
2810 		ret = -EBUSY;
2811 		if (!down_write_trylock(&F2FS_I(dst)->i_gc_rwsem[WRITE]))
2812 			goto out_src;
2813 	}
2814 
2815 	f2fs_lock_op(sbi);
2816 	ret = __exchange_data_block(src, dst, pos_in >> F2FS_BLKSIZE_BITS,
2817 				pos_out >> F2FS_BLKSIZE_BITS,
2818 				len >> F2FS_BLKSIZE_BITS, false);
2819 
2820 	if (!ret) {
2821 		if (dst_max_i_size)
2822 			f2fs_i_size_write(dst, dst_max_i_size);
2823 		else if (dst_osize != dst->i_size)
2824 			f2fs_i_size_write(dst, dst_osize);
2825 	}
2826 	f2fs_unlock_op(sbi);
2827 
2828 	if (src != dst)
2829 		up_write(&F2FS_I(dst)->i_gc_rwsem[WRITE]);
2830 out_src:
2831 	up_write(&F2FS_I(src)->i_gc_rwsem[WRITE]);
2832 out_unlock:
2833 	if (src != dst)
2834 		inode_unlock(dst);
2835 out:
2836 	inode_unlock(src);
2837 	return ret;
2838 }
2839 
2840 static int f2fs_ioc_move_range(struct file *filp, unsigned long arg)
2841 {
2842 	struct f2fs_move_range range;
2843 	struct fd dst;
2844 	int err;
2845 
2846 	if (!(filp->f_mode & FMODE_READ) ||
2847 			!(filp->f_mode & FMODE_WRITE))
2848 		return -EBADF;
2849 
2850 	if (copy_from_user(&range, (struct f2fs_move_range __user *)arg,
2851 							sizeof(range)))
2852 		return -EFAULT;
2853 
2854 	dst = fdget(range.dst_fd);
2855 	if (!dst.file)
2856 		return -EBADF;
2857 
2858 	if (!(dst.file->f_mode & FMODE_WRITE)) {
2859 		err = -EBADF;
2860 		goto err_out;
2861 	}
2862 
2863 	err = mnt_want_write_file(filp);
2864 	if (err)
2865 		goto err_out;
2866 
2867 	err = f2fs_move_file_range(filp, range.pos_in, dst.file,
2868 					range.pos_out, range.len);
2869 
2870 	mnt_drop_write_file(filp);
2871 	if (err)
2872 		goto err_out;
2873 
2874 	if (copy_to_user((struct f2fs_move_range __user *)arg,
2875 						&range, sizeof(range)))
2876 		err = -EFAULT;
2877 err_out:
2878 	fdput(dst);
2879 	return err;
2880 }
2881 
2882 static int f2fs_ioc_flush_device(struct file *filp, unsigned long arg)
2883 {
2884 	struct inode *inode = file_inode(filp);
2885 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2886 	struct sit_info *sm = SIT_I(sbi);
2887 	unsigned int start_segno = 0, end_segno = 0;
2888 	unsigned int dev_start_segno = 0, dev_end_segno = 0;
2889 	struct f2fs_flush_device range;
2890 	int ret;
2891 
2892 	if (!capable(CAP_SYS_ADMIN))
2893 		return -EPERM;
2894 
2895 	if (f2fs_readonly(sbi->sb))
2896 		return -EROFS;
2897 
2898 	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
2899 		return -EINVAL;
2900 
2901 	if (copy_from_user(&range, (struct f2fs_flush_device __user *)arg,
2902 							sizeof(range)))
2903 		return -EFAULT;
2904 
2905 	if (!f2fs_is_multi_device(sbi) || sbi->s_ndevs - 1 <= range.dev_num ||
2906 			__is_large_section(sbi)) {
2907 		f2fs_warn(sbi, "Can't flush %u in %d for segs_per_sec %u != 1",
2908 			  range.dev_num, sbi->s_ndevs, sbi->segs_per_sec);
2909 		return -EINVAL;
2910 	}
2911 
2912 	ret = mnt_want_write_file(filp);
2913 	if (ret)
2914 		return ret;
2915 
2916 	if (range.dev_num != 0)
2917 		dev_start_segno = GET_SEGNO(sbi, FDEV(range.dev_num).start_blk);
2918 	dev_end_segno = GET_SEGNO(sbi, FDEV(range.dev_num).end_blk);
2919 
2920 	start_segno = sm->last_victim[FLUSH_DEVICE];
2921 	if (start_segno < dev_start_segno || start_segno >= dev_end_segno)
2922 		start_segno = dev_start_segno;
2923 	end_segno = min(start_segno + range.segments, dev_end_segno);
2924 
2925 	while (start_segno < end_segno) {
2926 		if (!down_write_trylock(&sbi->gc_lock)) {
2927 			ret = -EBUSY;
2928 			goto out;
2929 		}
2930 		sm->last_victim[GC_CB] = end_segno + 1;
2931 		sm->last_victim[GC_GREEDY] = end_segno + 1;
2932 		sm->last_victim[ALLOC_NEXT] = end_segno + 1;
2933 		ret = f2fs_gc(sbi, true, true, start_segno);
2934 		if (ret == -EAGAIN)
2935 			ret = 0;
2936 		else if (ret < 0)
2937 			break;
2938 		start_segno++;
2939 	}
2940 out:
2941 	mnt_drop_write_file(filp);
2942 	return ret;
2943 }
2944 
2945 static int f2fs_ioc_get_features(struct file *filp, unsigned long arg)
2946 {
2947 	struct inode *inode = file_inode(filp);
2948 	u32 sb_feature = le32_to_cpu(F2FS_I_SB(inode)->raw_super->feature);
2949 
2950 	/* Must validate to set it with SQLite behavior in Android. */
2951 	sb_feature |= F2FS_FEATURE_ATOMIC_WRITE;
2952 
2953 	return put_user(sb_feature, (u32 __user *)arg);
2954 }
2955 
2956 #ifdef CONFIG_QUOTA
2957 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
2958 {
2959 	struct dquot *transfer_to[MAXQUOTAS] = {};
2960 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2961 	struct super_block *sb = sbi->sb;
2962 	int err = 0;
2963 
2964 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
2965 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
2966 		err = __dquot_transfer(inode, transfer_to);
2967 		if (err)
2968 			set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2969 		dqput(transfer_to[PRJQUOTA]);
2970 	}
2971 	return err;
2972 }
2973 
2974 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
2975 {
2976 	struct inode *inode = file_inode(filp);
2977 	struct f2fs_inode_info *fi = F2FS_I(inode);
2978 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2979 	struct page *ipage;
2980 	kprojid_t kprojid;
2981 	int err;
2982 
2983 	if (!f2fs_sb_has_project_quota(sbi)) {
2984 		if (projid != F2FS_DEF_PROJID)
2985 			return -EOPNOTSUPP;
2986 		else
2987 			return 0;
2988 	}
2989 
2990 	if (!f2fs_has_extra_attr(inode))
2991 		return -EOPNOTSUPP;
2992 
2993 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
2994 
2995 	if (projid_eq(kprojid, F2FS_I(inode)->i_projid))
2996 		return 0;
2997 
2998 	err = -EPERM;
2999 	/* Is it quota file? Do not allow user to mess with it */
3000 	if (IS_NOQUOTA(inode))
3001 		return err;
3002 
3003 	ipage = f2fs_get_node_page(sbi, inode->i_ino);
3004 	if (IS_ERR(ipage))
3005 		return PTR_ERR(ipage);
3006 
3007 	if (!F2FS_FITS_IN_INODE(F2FS_INODE(ipage), fi->i_extra_isize,
3008 								i_projid)) {
3009 		err = -EOVERFLOW;
3010 		f2fs_put_page(ipage, 1);
3011 		return err;
3012 	}
3013 	f2fs_put_page(ipage, 1);
3014 
3015 	err = dquot_initialize(inode);
3016 	if (err)
3017 		return err;
3018 
3019 	f2fs_lock_op(sbi);
3020 	err = f2fs_transfer_project_quota(inode, kprojid);
3021 	if (err)
3022 		goto out_unlock;
3023 
3024 	F2FS_I(inode)->i_projid = kprojid;
3025 	inode->i_ctime = current_time(inode);
3026 	f2fs_mark_inode_dirty_sync(inode, true);
3027 out_unlock:
3028 	f2fs_unlock_op(sbi);
3029 	return err;
3030 }
3031 #else
3032 int f2fs_transfer_project_quota(struct inode *inode, kprojid_t kprojid)
3033 {
3034 	return 0;
3035 }
3036 
3037 static int f2fs_ioc_setproject(struct file *filp, __u32 projid)
3038 {
3039 	if (projid != F2FS_DEF_PROJID)
3040 		return -EOPNOTSUPP;
3041 	return 0;
3042 }
3043 #endif
3044 
3045 /* FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR support */
3046 
3047 /*
3048  * To make a new on-disk f2fs i_flag gettable via FS_IOC_FSGETXATTR and settable
3049  * via FS_IOC_FSSETXATTR, add an entry for it to f2fs_xflags_map[], and add its
3050  * FS_XFLAG_* equivalent to F2FS_SUPPORTED_XFLAGS.
3051  */
3052 
3053 static const struct {
3054 	u32 iflag;
3055 	u32 xflag;
3056 } f2fs_xflags_map[] = {
3057 	{ F2FS_SYNC_FL,		FS_XFLAG_SYNC },
3058 	{ F2FS_IMMUTABLE_FL,	FS_XFLAG_IMMUTABLE },
3059 	{ F2FS_APPEND_FL,	FS_XFLAG_APPEND },
3060 	{ F2FS_NODUMP_FL,	FS_XFLAG_NODUMP },
3061 	{ F2FS_NOATIME_FL,	FS_XFLAG_NOATIME },
3062 	{ F2FS_PROJINHERIT_FL,	FS_XFLAG_PROJINHERIT },
3063 };
3064 
3065 #define F2FS_SUPPORTED_XFLAGS (		\
3066 		FS_XFLAG_SYNC |		\
3067 		FS_XFLAG_IMMUTABLE |	\
3068 		FS_XFLAG_APPEND |	\
3069 		FS_XFLAG_NODUMP |	\
3070 		FS_XFLAG_NOATIME |	\
3071 		FS_XFLAG_PROJINHERIT)
3072 
3073 /* Convert f2fs on-disk i_flags to FS_IOC_FS{GET,SET}XATTR flags */
3074 static inline u32 f2fs_iflags_to_xflags(u32 iflags)
3075 {
3076 	u32 xflags = 0;
3077 	int i;
3078 
3079 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3080 		if (iflags & f2fs_xflags_map[i].iflag)
3081 			xflags |= f2fs_xflags_map[i].xflag;
3082 
3083 	return xflags;
3084 }
3085 
3086 /* Convert FS_IOC_FS{GET,SET}XATTR flags to f2fs on-disk i_flags */
3087 static inline u32 f2fs_xflags_to_iflags(u32 xflags)
3088 {
3089 	u32 iflags = 0;
3090 	int i;
3091 
3092 	for (i = 0; i < ARRAY_SIZE(f2fs_xflags_map); i++)
3093 		if (xflags & f2fs_xflags_map[i].xflag)
3094 			iflags |= f2fs_xflags_map[i].iflag;
3095 
3096 	return iflags;
3097 }
3098 
3099 static void f2fs_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
3100 {
3101 	struct f2fs_inode_info *fi = F2FS_I(inode);
3102 
3103 	simple_fill_fsxattr(fa, f2fs_iflags_to_xflags(fi->i_flags));
3104 
3105 	if (f2fs_sb_has_project_quota(F2FS_I_SB(inode)))
3106 		fa->fsx_projid = from_kprojid(&init_user_ns, fi->i_projid);
3107 }
3108 
3109 static int f2fs_ioc_fsgetxattr(struct file *filp, unsigned long arg)
3110 {
3111 	struct inode *inode = file_inode(filp);
3112 	struct fsxattr fa;
3113 
3114 	f2fs_fill_fsxattr(inode, &fa);
3115 
3116 	if (copy_to_user((struct fsxattr __user *)arg, &fa, sizeof(fa)))
3117 		return -EFAULT;
3118 	return 0;
3119 }
3120 
3121 static int f2fs_ioc_fssetxattr(struct file *filp, unsigned long arg)
3122 {
3123 	struct inode *inode = file_inode(filp);
3124 	struct fsxattr fa, old_fa;
3125 	u32 iflags;
3126 	int err;
3127 
3128 	if (copy_from_user(&fa, (struct fsxattr __user *)arg, sizeof(fa)))
3129 		return -EFAULT;
3130 
3131 	/* Make sure caller has proper permission */
3132 	if (!inode_owner_or_capable(inode))
3133 		return -EACCES;
3134 
3135 	if (fa.fsx_xflags & ~F2FS_SUPPORTED_XFLAGS)
3136 		return -EOPNOTSUPP;
3137 
3138 	iflags = f2fs_xflags_to_iflags(fa.fsx_xflags);
3139 	if (f2fs_mask_flags(inode->i_mode, iflags) != iflags)
3140 		return -EOPNOTSUPP;
3141 
3142 	err = mnt_want_write_file(filp);
3143 	if (err)
3144 		return err;
3145 
3146 	inode_lock(inode);
3147 
3148 	f2fs_fill_fsxattr(inode, &old_fa);
3149 	err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
3150 	if (err)
3151 		goto out;
3152 
3153 	err = f2fs_setflags_common(inode, iflags,
3154 			f2fs_xflags_to_iflags(F2FS_SUPPORTED_XFLAGS));
3155 	if (err)
3156 		goto out;
3157 
3158 	err = f2fs_ioc_setproject(filp, fa.fsx_projid);
3159 out:
3160 	inode_unlock(inode);
3161 	mnt_drop_write_file(filp);
3162 	return err;
3163 }
3164 
3165 int f2fs_pin_file_control(struct inode *inode, bool inc)
3166 {
3167 	struct f2fs_inode_info *fi = F2FS_I(inode);
3168 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3169 
3170 	/* Use i_gc_failures for normal file as a risk signal. */
3171 	if (inc)
3172 		f2fs_i_gc_failures_write(inode,
3173 				fi->i_gc_failures[GC_FAILURE_PIN] + 1);
3174 
3175 	if (fi->i_gc_failures[GC_FAILURE_PIN] > sbi->gc_pin_file_threshold) {
3176 		f2fs_warn(sbi, "%s: Enable GC = ino %lx after %x GC trials",
3177 			  __func__, inode->i_ino,
3178 			  fi->i_gc_failures[GC_FAILURE_PIN]);
3179 		clear_inode_flag(inode, FI_PIN_FILE);
3180 		return -EAGAIN;
3181 	}
3182 	return 0;
3183 }
3184 
3185 static int f2fs_ioc_set_pin_file(struct file *filp, unsigned long arg)
3186 {
3187 	struct inode *inode = file_inode(filp);
3188 	__u32 pin;
3189 	int ret = 0;
3190 
3191 	if (get_user(pin, (__u32 __user *)arg))
3192 		return -EFAULT;
3193 
3194 	if (!S_ISREG(inode->i_mode))
3195 		return -EINVAL;
3196 
3197 	if (f2fs_readonly(F2FS_I_SB(inode)->sb))
3198 		return -EROFS;
3199 
3200 	ret = mnt_want_write_file(filp);
3201 	if (ret)
3202 		return ret;
3203 
3204 	inode_lock(inode);
3205 
3206 	if (f2fs_should_update_outplace(inode, NULL)) {
3207 		ret = -EINVAL;
3208 		goto out;
3209 	}
3210 
3211 	if (!pin) {
3212 		clear_inode_flag(inode, FI_PIN_FILE);
3213 		f2fs_i_gc_failures_write(inode, 0);
3214 		goto done;
3215 	}
3216 
3217 	if (f2fs_pin_file_control(inode, false)) {
3218 		ret = -EAGAIN;
3219 		goto out;
3220 	}
3221 
3222 	ret = f2fs_convert_inline_inode(inode);
3223 	if (ret)
3224 		goto out;
3225 
3226 	if (f2fs_disable_compressed_file(inode)) {
3227 		ret = -EOPNOTSUPP;
3228 		goto out;
3229 	}
3230 
3231 	set_inode_flag(inode, FI_PIN_FILE);
3232 	ret = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3233 done:
3234 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3235 out:
3236 	inode_unlock(inode);
3237 	mnt_drop_write_file(filp);
3238 	return ret;
3239 }
3240 
3241 static int f2fs_ioc_get_pin_file(struct file *filp, unsigned long arg)
3242 {
3243 	struct inode *inode = file_inode(filp);
3244 	__u32 pin = 0;
3245 
3246 	if (is_inode_flag_set(inode, FI_PIN_FILE))
3247 		pin = F2FS_I(inode)->i_gc_failures[GC_FAILURE_PIN];
3248 	return put_user(pin, (u32 __user *)arg);
3249 }
3250 
3251 int f2fs_precache_extents(struct inode *inode)
3252 {
3253 	struct f2fs_inode_info *fi = F2FS_I(inode);
3254 	struct f2fs_map_blocks map;
3255 	pgoff_t m_next_extent;
3256 	loff_t end;
3257 	int err;
3258 
3259 	if (is_inode_flag_set(inode, FI_NO_EXTENT))
3260 		return -EOPNOTSUPP;
3261 
3262 	map.m_lblk = 0;
3263 	map.m_next_pgofs = NULL;
3264 	map.m_next_extent = &m_next_extent;
3265 	map.m_seg_type = NO_CHECK_TYPE;
3266 	map.m_may_create = false;
3267 	end = F2FS_I_SB(inode)->max_file_blocks;
3268 
3269 	while (map.m_lblk < end) {
3270 		map.m_len = end - map.m_lblk;
3271 
3272 		down_write(&fi->i_gc_rwsem[WRITE]);
3273 		err = f2fs_map_blocks(inode, &map, 0, F2FS_GET_BLOCK_PRECACHE);
3274 		up_write(&fi->i_gc_rwsem[WRITE]);
3275 		if (err)
3276 			return err;
3277 
3278 		map.m_lblk = m_next_extent;
3279 	}
3280 
3281 	return err;
3282 }
3283 
3284 static int f2fs_ioc_precache_extents(struct file *filp, unsigned long arg)
3285 {
3286 	return f2fs_precache_extents(file_inode(filp));
3287 }
3288 
3289 static int f2fs_ioc_resize_fs(struct file *filp, unsigned long arg)
3290 {
3291 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
3292 	__u64 block_count;
3293 	int ret;
3294 
3295 	if (!capable(CAP_SYS_ADMIN))
3296 		return -EPERM;
3297 
3298 	if (f2fs_readonly(sbi->sb))
3299 		return -EROFS;
3300 
3301 	if (copy_from_user(&block_count, (void __user *)arg,
3302 			   sizeof(block_count)))
3303 		return -EFAULT;
3304 
3305 	ret = f2fs_resize_fs(sbi, block_count);
3306 
3307 	return ret;
3308 }
3309 
3310 static int f2fs_ioc_enable_verity(struct file *filp, unsigned long arg)
3311 {
3312 	struct inode *inode = file_inode(filp);
3313 
3314 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
3315 
3316 	if (!f2fs_sb_has_verity(F2FS_I_SB(inode))) {
3317 		f2fs_warn(F2FS_I_SB(inode),
3318 			  "Can't enable fs-verity on inode %lu: the verity feature is not enabled on this filesystem.\n",
3319 			  inode->i_ino);
3320 		return -EOPNOTSUPP;
3321 	}
3322 
3323 	return fsverity_ioctl_enable(filp, (const void __user *)arg);
3324 }
3325 
3326 static int f2fs_ioc_measure_verity(struct file *filp, unsigned long arg)
3327 {
3328 	if (!f2fs_sb_has_verity(F2FS_I_SB(file_inode(filp))))
3329 		return -EOPNOTSUPP;
3330 
3331 	return fsverity_ioctl_measure(filp, (void __user *)arg);
3332 }
3333 
3334 static int f2fs_get_volume_name(struct file *filp, unsigned long arg)
3335 {
3336 	struct inode *inode = file_inode(filp);
3337 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3338 	char *vbuf;
3339 	int count;
3340 	int err = 0;
3341 
3342 	vbuf = f2fs_kzalloc(sbi, MAX_VOLUME_NAME, GFP_KERNEL);
3343 	if (!vbuf)
3344 		return -ENOMEM;
3345 
3346 	down_read(&sbi->sb_lock);
3347 	count = utf16s_to_utf8s(sbi->raw_super->volume_name,
3348 			ARRAY_SIZE(sbi->raw_super->volume_name),
3349 			UTF16_LITTLE_ENDIAN, vbuf, MAX_VOLUME_NAME);
3350 	up_read(&sbi->sb_lock);
3351 
3352 	if (copy_to_user((char __user *)arg, vbuf,
3353 				min(FSLABEL_MAX, count)))
3354 		err = -EFAULT;
3355 
3356 	kvfree(vbuf);
3357 	return err;
3358 }
3359 
3360 static int f2fs_set_volume_name(struct file *filp, unsigned long arg)
3361 {
3362 	struct inode *inode = file_inode(filp);
3363 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
3364 	char *vbuf;
3365 	int err = 0;
3366 
3367 	if (!capable(CAP_SYS_ADMIN))
3368 		return -EPERM;
3369 
3370 	vbuf = strndup_user((const char __user *)arg, FSLABEL_MAX);
3371 	if (IS_ERR(vbuf))
3372 		return PTR_ERR(vbuf);
3373 
3374 	err = mnt_want_write_file(filp);
3375 	if (err)
3376 		goto out;
3377 
3378 	down_write(&sbi->sb_lock);
3379 
3380 	memset(sbi->raw_super->volume_name, 0,
3381 			sizeof(sbi->raw_super->volume_name));
3382 	utf8s_to_utf16s(vbuf, strlen(vbuf), UTF16_LITTLE_ENDIAN,
3383 			sbi->raw_super->volume_name,
3384 			ARRAY_SIZE(sbi->raw_super->volume_name));
3385 
3386 	err = f2fs_commit_super(sbi, false);
3387 
3388 	up_write(&sbi->sb_lock);
3389 
3390 	mnt_drop_write_file(filp);
3391 out:
3392 	kfree(vbuf);
3393 	return err;
3394 }
3395 
3396 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
3397 {
3398 	if (unlikely(f2fs_cp_error(F2FS_I_SB(file_inode(filp)))))
3399 		return -EIO;
3400 	if (!f2fs_is_checkpoint_ready(F2FS_I_SB(file_inode(filp))))
3401 		return -ENOSPC;
3402 
3403 	switch (cmd) {
3404 	case F2FS_IOC_GETFLAGS:
3405 		return f2fs_ioc_getflags(filp, arg);
3406 	case F2FS_IOC_SETFLAGS:
3407 		return f2fs_ioc_setflags(filp, arg);
3408 	case F2FS_IOC_GETVERSION:
3409 		return f2fs_ioc_getversion(filp, arg);
3410 	case F2FS_IOC_START_ATOMIC_WRITE:
3411 		return f2fs_ioc_start_atomic_write(filp);
3412 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3413 		return f2fs_ioc_commit_atomic_write(filp);
3414 	case F2FS_IOC_START_VOLATILE_WRITE:
3415 		return f2fs_ioc_start_volatile_write(filp);
3416 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3417 		return f2fs_ioc_release_volatile_write(filp);
3418 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3419 		return f2fs_ioc_abort_volatile_write(filp);
3420 	case F2FS_IOC_SHUTDOWN:
3421 		return f2fs_ioc_shutdown(filp, arg);
3422 	case FITRIM:
3423 		return f2fs_ioc_fitrim(filp, arg);
3424 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3425 		return f2fs_ioc_set_encryption_policy(filp, arg);
3426 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3427 		return f2fs_ioc_get_encryption_policy(filp, arg);
3428 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3429 		return f2fs_ioc_get_encryption_pwsalt(filp, arg);
3430 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3431 		return f2fs_ioc_get_encryption_policy_ex(filp, arg);
3432 	case FS_IOC_ADD_ENCRYPTION_KEY:
3433 		return f2fs_ioc_add_encryption_key(filp, arg);
3434 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3435 		return f2fs_ioc_remove_encryption_key(filp, arg);
3436 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3437 		return f2fs_ioc_remove_encryption_key_all_users(filp, arg);
3438 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3439 		return f2fs_ioc_get_encryption_key_status(filp, arg);
3440 	case F2FS_IOC_GARBAGE_COLLECT:
3441 		return f2fs_ioc_gc(filp, arg);
3442 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3443 		return f2fs_ioc_gc_range(filp, arg);
3444 	case F2FS_IOC_WRITE_CHECKPOINT:
3445 		return f2fs_ioc_write_checkpoint(filp, arg);
3446 	case F2FS_IOC_DEFRAGMENT:
3447 		return f2fs_ioc_defragment(filp, arg);
3448 	case F2FS_IOC_MOVE_RANGE:
3449 		return f2fs_ioc_move_range(filp, arg);
3450 	case F2FS_IOC_FLUSH_DEVICE:
3451 		return f2fs_ioc_flush_device(filp, arg);
3452 	case F2FS_IOC_GET_FEATURES:
3453 		return f2fs_ioc_get_features(filp, arg);
3454 	case F2FS_IOC_FSGETXATTR:
3455 		return f2fs_ioc_fsgetxattr(filp, arg);
3456 	case F2FS_IOC_FSSETXATTR:
3457 		return f2fs_ioc_fssetxattr(filp, arg);
3458 	case F2FS_IOC_GET_PIN_FILE:
3459 		return f2fs_ioc_get_pin_file(filp, arg);
3460 	case F2FS_IOC_SET_PIN_FILE:
3461 		return f2fs_ioc_set_pin_file(filp, arg);
3462 	case F2FS_IOC_PRECACHE_EXTENTS:
3463 		return f2fs_ioc_precache_extents(filp, arg);
3464 	case F2FS_IOC_RESIZE_FS:
3465 		return f2fs_ioc_resize_fs(filp, arg);
3466 	case FS_IOC_ENABLE_VERITY:
3467 		return f2fs_ioc_enable_verity(filp, arg);
3468 	case FS_IOC_MEASURE_VERITY:
3469 		return f2fs_ioc_measure_verity(filp, arg);
3470 	case F2FS_IOC_GET_VOLUME_NAME:
3471 		return f2fs_get_volume_name(filp, arg);
3472 	case F2FS_IOC_SET_VOLUME_NAME:
3473 		return f2fs_set_volume_name(filp, arg);
3474 	default:
3475 		return -ENOTTY;
3476 	}
3477 }
3478 
3479 static ssize_t f2fs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
3480 {
3481 	struct file *file = iocb->ki_filp;
3482 	struct inode *inode = file_inode(file);
3483 
3484 	if (!f2fs_is_compress_backend_ready(inode))
3485 		return -EOPNOTSUPP;
3486 
3487 	return generic_file_read_iter(iocb, iter);
3488 }
3489 
3490 static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
3491 {
3492 	struct file *file = iocb->ki_filp;
3493 	struct inode *inode = file_inode(file);
3494 	ssize_t ret;
3495 
3496 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
3497 		ret = -EIO;
3498 		goto out;
3499 	}
3500 
3501 	if (!f2fs_is_compress_backend_ready(inode))
3502 		return -EOPNOTSUPP;
3503 
3504 	if (iocb->ki_flags & IOCB_NOWAIT) {
3505 		if (!inode_trylock(inode)) {
3506 			ret = -EAGAIN;
3507 			goto out;
3508 		}
3509 	} else {
3510 		inode_lock(inode);
3511 	}
3512 
3513 	ret = generic_write_checks(iocb, from);
3514 	if (ret > 0) {
3515 		bool preallocated = false;
3516 		size_t target_size = 0;
3517 		int err;
3518 
3519 		if (iov_iter_fault_in_readable(from, iov_iter_count(from)))
3520 			set_inode_flag(inode, FI_NO_PREALLOC);
3521 
3522 		if ((iocb->ki_flags & IOCB_NOWAIT)) {
3523 			if (!f2fs_overwrite_io(inode, iocb->ki_pos,
3524 						iov_iter_count(from)) ||
3525 				f2fs_has_inline_data(inode) ||
3526 				f2fs_force_buffered_io(inode, iocb, from)) {
3527 				clear_inode_flag(inode, FI_NO_PREALLOC);
3528 				inode_unlock(inode);
3529 				ret = -EAGAIN;
3530 				goto out;
3531 			}
3532 			goto write;
3533 		}
3534 
3535 		if (is_inode_flag_set(inode, FI_NO_PREALLOC))
3536 			goto write;
3537 
3538 		if (iocb->ki_flags & IOCB_DIRECT) {
3539 			/*
3540 			 * Convert inline data for Direct I/O before entering
3541 			 * f2fs_direct_IO().
3542 			 */
3543 			err = f2fs_convert_inline_inode(inode);
3544 			if (err)
3545 				goto out_err;
3546 			/*
3547 			 * If force_buffere_io() is true, we have to allocate
3548 			 * blocks all the time, since f2fs_direct_IO will fall
3549 			 * back to buffered IO.
3550 			 */
3551 			if (!f2fs_force_buffered_io(inode, iocb, from) &&
3552 					allow_outplace_dio(inode, iocb, from))
3553 				goto write;
3554 		}
3555 		preallocated = true;
3556 		target_size = iocb->ki_pos + iov_iter_count(from);
3557 
3558 		err = f2fs_preallocate_blocks(iocb, from);
3559 		if (err) {
3560 out_err:
3561 			clear_inode_flag(inode, FI_NO_PREALLOC);
3562 			inode_unlock(inode);
3563 			ret = err;
3564 			goto out;
3565 		}
3566 write:
3567 		ret = __generic_file_write_iter(iocb, from);
3568 		clear_inode_flag(inode, FI_NO_PREALLOC);
3569 
3570 		/* if we couldn't write data, we should deallocate blocks. */
3571 		if (preallocated && i_size_read(inode) < target_size)
3572 			f2fs_truncate(inode);
3573 
3574 		if (ret > 0)
3575 			f2fs_update_iostat(F2FS_I_SB(inode), APP_WRITE_IO, ret);
3576 	}
3577 	inode_unlock(inode);
3578 out:
3579 	trace_f2fs_file_write_iter(inode, iocb->ki_pos,
3580 					iov_iter_count(from), ret);
3581 	if (ret > 0)
3582 		ret = generic_write_sync(iocb, ret);
3583 	return ret;
3584 }
3585 
3586 #ifdef CONFIG_COMPAT
3587 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
3588 {
3589 	switch (cmd) {
3590 	case F2FS_IOC32_GETFLAGS:
3591 		cmd = F2FS_IOC_GETFLAGS;
3592 		break;
3593 	case F2FS_IOC32_SETFLAGS:
3594 		cmd = F2FS_IOC_SETFLAGS;
3595 		break;
3596 	case F2FS_IOC32_GETVERSION:
3597 		cmd = F2FS_IOC_GETVERSION;
3598 		break;
3599 	case F2FS_IOC_START_ATOMIC_WRITE:
3600 	case F2FS_IOC_COMMIT_ATOMIC_WRITE:
3601 	case F2FS_IOC_START_VOLATILE_WRITE:
3602 	case F2FS_IOC_RELEASE_VOLATILE_WRITE:
3603 	case F2FS_IOC_ABORT_VOLATILE_WRITE:
3604 	case F2FS_IOC_SHUTDOWN:
3605 	case FITRIM:
3606 	case F2FS_IOC_SET_ENCRYPTION_POLICY:
3607 	case F2FS_IOC_GET_ENCRYPTION_PWSALT:
3608 	case F2FS_IOC_GET_ENCRYPTION_POLICY:
3609 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
3610 	case FS_IOC_ADD_ENCRYPTION_KEY:
3611 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
3612 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
3613 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
3614 	case F2FS_IOC_GARBAGE_COLLECT:
3615 	case F2FS_IOC_GARBAGE_COLLECT_RANGE:
3616 	case F2FS_IOC_WRITE_CHECKPOINT:
3617 	case F2FS_IOC_DEFRAGMENT:
3618 	case F2FS_IOC_MOVE_RANGE:
3619 	case F2FS_IOC_FLUSH_DEVICE:
3620 	case F2FS_IOC_GET_FEATURES:
3621 	case F2FS_IOC_FSGETXATTR:
3622 	case F2FS_IOC_FSSETXATTR:
3623 	case F2FS_IOC_GET_PIN_FILE:
3624 	case F2FS_IOC_SET_PIN_FILE:
3625 	case F2FS_IOC_PRECACHE_EXTENTS:
3626 	case F2FS_IOC_RESIZE_FS:
3627 	case FS_IOC_ENABLE_VERITY:
3628 	case FS_IOC_MEASURE_VERITY:
3629 	case F2FS_IOC_GET_VOLUME_NAME:
3630 	case F2FS_IOC_SET_VOLUME_NAME:
3631 		break;
3632 	default:
3633 		return -ENOIOCTLCMD;
3634 	}
3635 	return f2fs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
3636 }
3637 #endif
3638 
3639 const struct file_operations f2fs_file_operations = {
3640 	.llseek		= f2fs_llseek,
3641 	.read_iter	= f2fs_file_read_iter,
3642 	.write_iter	= f2fs_file_write_iter,
3643 	.open		= f2fs_file_open,
3644 	.release	= f2fs_release_file,
3645 	.mmap		= f2fs_file_mmap,
3646 	.flush		= f2fs_file_flush,
3647 	.fsync		= f2fs_sync_file,
3648 	.fallocate	= f2fs_fallocate,
3649 	.unlocked_ioctl	= f2fs_ioctl,
3650 #ifdef CONFIG_COMPAT
3651 	.compat_ioctl	= f2fs_compat_ioctl,
3652 #endif
3653 	.splice_read	= generic_file_splice_read,
3654 	.splice_write	= iter_file_splice_write,
3655 };
3656